AU2024311097A1 - Modeling a user predisposition based on location data - Google Patents

Abstract

A method and system for modeling a user predisposition based on location data are provided. The method involves training at least one model to determine a user's predisposition for a particular behavior using location data for multiple users, inferring activity data from location data, and translating activity data and location data into time- varying and static behavior attributes for each user. The trained model is then used to predict predisposition for a particular behavior for an individual user by obtaining their location data, inputting it into the trained model, and receiving an assigned quantified predisposition for the particular behavior. The method can be applied to various behaviors such as purchase intent, hospitalization risk, job participation, job change, travel intent, residential relocation intent, and healthcare risk.

Description

PATENT APPLICATION F^OR MODELING A USER PREDISPOSITION BASED ON LOCATION DATA BY M^EGHANATH M^ACHA BEIBEI LI NASEER HASHIM CROSS-REFERENCE TO RELATED APPLICATION(S) [0001] This application claims priority to and the benefit of co-pending United States Provisional Application 63/523,874 filed June 28, 2023, for all subject matter common to both applications. The disclosure of said provisional application is hereby incorporated by reference in its entirety. FIELD OF THE INVENTION [0002] The present invention relates to the field of data analysis and machine learning, suitable for behavior modeling. In particular, the present invention relates to modeling user predispositions based on location data. BACKGROUND [0003] In recent years, the ubiquity of mobile devices and the rapid development of location-based services have led to exponential growth in the availability of location data. This data, which includes information about users' geographical positions and movements, has opened new possibilities for understanding and predicting human behavior. One area of interest is the modeling of user predispositions based on location data, which can provide valuable insights into users' preferences, habits, and potential future actions or experiences. [0004] Traditional methods of modeling user predispositions have often relied on self- reported data, such as surveys and questionnaires. However, these methods can be subject to various biases and inaccuracies, as users may not accurately recall or report their past 4868-3590-9304, v.1 behaviors and experiences. Moreover, self-reported data may not capture the full range of factors that influence users’ predispositions, such as environmental and contextual factors that can be derived from location data. [0005] Machine learning techniques have been increasingly employed to model user predispositions based on location data. These techniques can automatically learn patterns and relationships in the data, enabling the development of more accurate and robust models. However, the process of training machine learning models can be complex and computationally intensive, requiring the careful selection and tuning of model parameters to achieve optimal performance. Furthermore, current approaches can result in an inflexible model that makes it difficult to repurpose or adapt the model to other uses. [0006] Furthermore, the integration of time-varying and static behavior attributes in the modeling process can be challenging. Time-varying attributes, such as the frequency and duration of visits to specific locations, can change over time and may be influenced by various factors, such as users' schedules and preferences. Static attributes, on the other hand, represent more stable characteristics of users, such as their demographic information or home location. Combining these different types of attributes in a meaningful and effective way is crucial for developing accurate models of user predispositions. S^UMMARY [0007] In light of these challenges, there is a need for a system to determine user predispositions based on location data that can effectively integrate time-varying and static behavior attributes, while also leveraging the power of machine learning techniques to automatically learn patterns and relationships in the data. The present invention enables the development of more accurate and robust systems to predict users' predispositions for particular behaviors or experiences/conditions with greater precision and reliability while maintaining the ability to adapt or repurpose the model to predict different behaviors, experiences, or behaviors as needed in a far more efficient manner. [0008] The present invention addresses this need by providing a method of determining user predispositions based on location data, which includes the steps of training at least one model to determine a user's predisposition for a particular behavior or experience/condition, and predicting predisposition for a particular behavior or experience/condition for an individual user. The method involves obtaining location data for multiple users, formatting 4868-3590-9304, v.1 the location data into trajectories, inferring activity data from the location data, translating the activity data and location data into time-varying and static behavior attributes, and determining predispositions for particular behaviors or experiences/conditions by modeling time-varying attributes, combining modeled time-varying attributes with static attributes, and assigning quantified predispositions. The method also includes adjusting model parameters based on results, resulting in a trained model, and using the trained model to predict predispositions for individual users based on their location data. [0009] Location data can be collected from various sources, such as GPS-enabled devices, Wi-Fi access points, and cell towers. This data can be used to create trajectories, which are sequences of geographical positions and timestamps that represent the movement of users over time. By analyzing these trajectories, it is possible to infer various types of activity data, such as the types of places visited, the duration of visits, and the frequency of visits to specific locations. [0010] The analysis of location data and activity data can reveal patterns and trends in users' behavior, which can be used to model their predispositions for particular behaviors or experiences/conditions. These predispositions can be quantified and used to predict the likelihood of users engaging in certain activities or experiencing specific conditions in the future. Such predictions can be valuable for various applications, including personalized recommendations, targeted advertising, and public health interventions. [0011] In accordance with certain embodiments, the techniques described herein relate to a method of modeling a user predisposition based on location data, the method including: I. training at least one model to determine a user's predisposition for a particular behavior or experience/condition, the training including: A) obtaining location data for multiple users; B) formatting location data into one or more trajectories for each user of multiple users; C) inferring activity data from location data; D) translating activity data and location data into time-varying and static behavior attributes for each user of the multiple users; E) determining predisposition for a particular behavior or experience/condition including: for each user: modeling time-varying attributes; combining modeled time-varying attributes with static attributes; and assigning a quantified predisposition for a particular behavior or experience/condition; and adjusting parameters based on results resulting in a trained model; II. predicting predisposition for a particular behavior or experience/condition for an individual user including: A) obtaining location data for the individual user; B) inputting the location data into the at least one trained model produced by the training step; and C) 4868-3590-9304, v.1 receiving an assigned quantified predisposition for a particular behavior or experience/condition for the individual user from the trained model. [0012] In accordance with certain aspects, the predisposition for a particular behavior or experience/condition is purchase intent; the predisposition for a particular behavior or experience/condition is hospitalization risk; the predisposition for a particular behavior or experience/condition is job participation; the predisposition for a particular behavior or experience/condition is job change; the predisposition for a particular behavior or experience/condition is travel intent; the predisposition for a particular behavior or experience/condition is residential relocation intent; the predisposition for a particular behavior or experience/condition is healthcare risk; and/or the predisposition for a particular behavior or experience/condition is associated with an opportunity window. [0013] In accordance with certain aspects, the location data is geospatial data for a user hardware device over a period of time. In accordance with certain aspects, the location data for a user device is provided by a location provider/vendor. In accordance with certain aspects, the location data for a user is identified without using one or more of: personal identifying information (PII), demographic information, and socioeconomic information about the user. [0014] In accordance with certain aspects, inferring activity data from location data further includes: i) mapping location data to place types; ii) grouping place types into activity groups based on a function of the place type; and iii) transforming the one or more trajectories for the user into one or more activity-trajectories for the user using the activity groups. In some such aspects, the activity groups include one or more selected from the group including: hospital, health, necessity shopping, fitness, public transport, own transport, religious, recreation, travel, personal care, leisure shopping, unhealthy activities, restaurant, home, and work. [0015] In accordance with certain aspects, translating activity data and location data into behavior attributes for each user of the multiple users includes: defining time-varying attributes; and defining static attributes. [0016] In some such aspects, defining time-varying attributes includes: determining a lifestyle attribute; determining activity attributes; and determining mobility attributes. In further such aspects, determining a lifestyle attribute includes using an unsupervised learning model that can identify similarities in activity patterns among users. In still further aspects, the unsupervised learning model includes a clustering and dimension reduction model, a 4868-3590-9304, v.1 Hidden Markov model, or an LDA and topic model. The unsupervised learning model includes an LDA and topic model may further include an Author topic models (ATM) where users are authors, activities are words, periods of activity are documents, and lifestyles are topics. [0017] In other such aspects, defining static attributes includes: determining accessibility attributes; and determining social demographic attributes. [0018] In still other such aspects, defining time-varying attributes includes: embedding the lifestyle attribute into a continuous vector representation of a fixed dimension having learnable weigh parameters and tunable model hyperparameters; concatenating the embedded lifestyle attribute and the attributes for the activity and mobility; and transforming the concatenated embedded lifestyle attribute and time varying numerical attributes into a hidden representation having shared learnable weight and bias parameters and a tunable model hyperparameter. In further still other aspects, transforming the concatenated embedded lifestyle attribute and time varying numerical attributes into a hidden representation includes using a supervised machine learning model that models both spatial and temporal information from a user trajectory. In still further still other aspects, the supervised machine learning model includes a non-deep-learning regression or classification model. The non-deep- learning regression or classification model can include a decision tree based model, a random forest based model, or a gradient boosting model. Alternately, the supervised machine learning model includes a deep learning model or a deep neural network based model. The deep neural network based model may includes at least one of: Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), Long Short Term Memory (LSTM) models, Radial Basis Function Networks (RBFN), Transformer-based attention models. The deep neural network based model may also include a Convoluted Long Short Term Memory (CLSTM) model with the time varying attributes as inputs. [0019] In further such aspects, combining modeled time-varying attributes with static attributes includes: embedding categorical static attributes into a continuous vector representation of a fixed dimension having learnable weigh parameters and tunable model hyperparameters; and concatenating the embedded categorical static attributes, numerical static attributes, and a hidden representation of the modeled time-varying attributes. [0020] In accordance with certain aspects, the model is tuned via cross-validation. [0021] In accordance with certain aspects, the method further includes outputting the assigned quantified predisposition for a particular behavior or experience/condition for the 4868-3590-9304, v.1 particular user. In some such aspects, the outputting includes providing a graphical representation of the assigned quantified predisposition for a particular behavior or experience/condition for the particular user. In other such aspects, the outputting includes providing a web service API providing access to the assigned quantified predisposition for a particular behavior or experience/condition for the particular user. [0022] In accordance with certain embodiments, a system for modeling user predisposition based on location data is provided. The system includes: a data collection module configured to obtain location data for multiple users; a model training module configured to train a model to determine a user's predisposition for a particular behavior using the location data; and a prediction module configured to predict predisposition for a particular behavior for an individual user using the trained model and location data for the individual user. [0023] In accordance with certain aspects, the model training module is further configured to infer activity data from location data, translate activity data and location data into behavior attributes, and determine predisposition for a particular behavior for each user. [0024] In accordance with certain aspects, the prediction module is further configured to input the location data into the trained model and receive an assigned quantified predisposition for a particular behavior for the particular user from the trained model. [0025] The present invention provides several advantages over previous approaches. The present invention provides an accurate prediction for a particular user instead of a prediction for a type or group of users based on group demographics. The present invention provides a flexible model that can be adapted or repurposed to predict different behaviors, experiences, or conditions as needed in a far more efficient manner. That is, the iterative training of the model allows the system to be tuned to provide for an accurate prediction for particular behaviors, experiences, or conditions and then be quickly and conveniently repurposed and retrained to predict a different behavior, experience, or condition all while using the same location data source (broadly) as an input. Furthermore, the present invention can also use publicly available location data to infer activity for users and as such does not require Personal Identifying Information (PII) about the users for the predictions. 4868-3590-9304, v.1 BRIEF DESCRIPTION OF THE FIGURES [0026] These and other characteristics of the present invention will be more fully understood by reference to the following detailed description in conjunction with the attached drawings, in which: [0027] FIG. 1 depicts an illustrative environment in which the invention is utilized in accordance with embodiments of the present invention; [0028] FIG. 2 is an illustrative flow chart depicting the process for modeling a user predisposition based on location data in accordance with embodiments of the present invention; [0029] FIG. 3 is an illustrative flow chart depicting the process involved in inferring activity data in accordance with embodiments of the present invention; [0030] FIG. 4 is an illustrative flow chart depicting the process involved in translating activity data and location data into behavior attributes in accordance with embodiments of the present invention; [0031] FIG. 5 is an illustrative flow chart depicting the process involved determining a predisposition and tuning in accordance with embodiments of the present invention; [0032] FIG. 6 is an illustrative dashboard graphically showing user predisposition in a region in accordance with embodiments of the present invention; [0033] FIG. 7 is an illustrative dashboard graphically showing user predisposition in a specific region in accordance with embodiments of the present invention; [0034] FIG. 8 is an illustrative dashboard graphically showing real-time predisposition modeling for a user in accordance with embodiments of the present invention; [0035] FIG. 9 is an illustrative dashboard graphically showing potential uses for such real-time in accordance with embodiments of the present invention; [0036] FIG. 10 is a diagrammatic illustration of a high-level architecture for implementing processes in accordance with embodiments of the present invention; [0037] FIG. 11 is an illustrative probabilistic graphical model of an Author Topic Model using plate notation in accordance with embodiments of the present invention; [0038] FIG. 12 is an illustrative architecture diagram of a proposed sequential deep- learning model in accordance with embodiments of the present invention; [0039] FIG. 13 is an illustrative graphical model of CLSTM and concatenate layers in accordance with embodiments of the present invention; and 4868-3590-9304, v.1 [0040] FIG. 14 is an illustrative heatmap of user activity in accordance with embodiments of the present invention. DETAILED DESCRIPTION [0041] An illustrative embodiment of the present invention relates to a technology that leverages large amounts of crowd-sourced location data to identify lifestyles and quantify their association with future behavior or experiences. When analyzing large amounts of location data from hundreds of thousands to millions of users, a number of key findings emerge. An individual's lifestyle choice is a more critical predictor of future behavior and experiences than other factors commonly believed to be heavily influential (like accessibility to healthcare, socio-economic factors, age, hobbies, etc.). For example, people with busy, varying, work routines and limited gym visits are 2.01 times more likely to be hospitalized within a year when compared to the population average. The technology enables predictions of a person’s predisposition toward certain behaviors or experiences. It leverages crowd- sourced location data and information indicating a location history for a large number of users, and then correlates the data to various behaviors and experiences those users have had. The correlations are then relied upon to predict future behaviors or experiences of a particular person based on their real-time and/or recent location history. Applications for this technology are vast. It can be applied to various behaviors or experiences such as a person’s intention to purchase a product, require hospitalization, go through job changes, embark on certain travel adventures, relocate their home, require certain healthcare, and other behaviors or experiences. [0042] FIG. 1 through FIG. 14, wherein like parts are designated by like reference numerals throughout, illustrate an example embodiment or embodiments of method and system for modeling a user predisposition based on location data, according to the present invention. Although the present invention will be described with reference to the example embodiment or embodiments illustrated in the figures, it should be understood that many alternative forms can embody the present invention. One of skill in the art will additionally appreciate different ways to alter the parameters of the embodiment(s) disclosed, such as the size, shape, or type of elements or materials, in a manner still in keeping with the spirit and scope of the present invention. 4868-3590-9304, v.1 [0043] FIG. 1 depicts an illustrative environment 100 in which the invention is utilized. The environment 100 includes multiple users 102, a server or cloud service 104, and a business 106. Each user 102 has a hardware or mobile device 108, such as a cell phone, that is connected to a network or internet 110. The server or cloud service 104 is a system having a data collection module 112 configured to obtain location data for multiple users, a model training module 114 configured to train a model to determine a user's predisposition for a particular behavior using the location data, and a prediction module 116 configured to predict predisposition for a particular behavior for an individual user using the trained model and location data for the individual user receives location data for the multiple users 102 based their respective hardware or mobile device 108 which can be located using GPS or network connectivity. In some embodiments, the location data is gathered and provided to the server or cloud service 104 by a vendor 118 who collects and sells location data. Using the received location data for the users 102 the server or cloud service 104 models behavior for the users and provides the business 106 a prediction for a user 102 behavior. The business can be any type of entity that relies on customers for their business models, such as retail businesses, hospitals, etc. The business can then tailor is advertising, marketing, of provided experience based on this prediction. [0044] FIG. 2 illustrates a flowchart 200 representing the various steps involved in the method for modeling user predisposition based on location data, as described in the patent claims. The flowchart consists of the following elements and their respective functions: [0045] Training the model (step 202) involves training at least one model to determine a user's predisposition for a particular behavior. The training process includes obtaining location data for multiple users (step 206), formatting the data into trajectories (step 208), inferring activity data (step 210), translating activity data and location data into behavior attributes (Step 212), and determining predisposition for a particular behavior or experience/condition (Step 214). [0046] In certain embodiments, the location data for the user is the raw geospatial data collected from user hardware devices, such as a mobile device 108, over a period of time. In some such embodiments, the location data is provided by a location provider/vendor 118. In some embodiments, the location data is identified without using personal identifying information (PII), demographic information, or socioeconomic information about the user 102. 4868-3590-9304, v.1 [0047] Formatting location data (Step 208) involves processing the collected location data into one or more trajectories for each user. In certain embodiments, a trajectory can be a mathematical representation. For example, a trajectory ^_^ of an individual ^ can be defined as a temporally ordered set of tuples is a location where ^_^ ^{^} and ^_^ ^{^} are the coordinates of the geographic location, and ^_^ ^{^} is the corresponding time stamp. Here, the individual ^ corresponds to a user 102. [0048] Inferring activity data (Step 210) involves mapping the geospatial location data to place types, grouping place types into activity groups, and transforming the trajectories into activity-trajectories for each user. FIG. 3 depicts the steps involved in Inferring Activity Data (Step 210). It includes mapping the geo-spatial location data to place types (Step 300), grouping place types into activity groups based on a function of the place type (Step 302), and transforming the one or more trajectories for the user into one or more activity- trajectories for the user using the activity groups (Step 304). [0049] Mapping location data to place types (Step 300) can involve mapping the collected geo-spatial location data to specific place types, such as hospitals, restaurants, or shopping centers. [0050] Grouping place types into activity groups (Step 302) can involve the place types being grouped into activity groups based on the function of the place type. Examples of activity groups include: hospital, health, necessity shopping, fitness, public transport, own transport, religious, recreation, travel, personal care, leisure shopping, unhealthy activities, restaurant, home, and work. [0051] Transforming the one or more trajectories for the user into one or more activity- trajectories for the user using the activity groups (Step 304) involves can involve a mathematical representation. For example, an activity trajectory ^_^ of an individual ^ can be defined as mapping ^_^ to activities that exhibit a pattern of behavior, consumption, leisure. ^_^is a temporally ordered set of tuples ^_^ = ^^_^ ^{^} , , … , ^_^^ ^^ , ^_^ ^{^} = ^^_^ ^{^}, ^_^ ^{^}^ where ^_^ ^{^} = ^^^^^_^ ^{^}^ ^_^ ^{^} ∈ ^_^ is an activity by the individual inferred from a location closest to ^_^ ^{^} and ^_^ ^{^}, and ^_^ ^{^} is a coarser timestamp of ^_^ ^{^}. Also, denote ^ as the universe of all temporal activities ^_^ ^{^} across _. This step allows for a more detailed analysis of user behavior based on their activities rather than just their location data. 4868-3590-9304, v.1 [0052] FIG. 4. depicts the steps involved in Translating Activity Data and Location Data into behaviors attributes for each user of the multiple users (Step 212). It involves defining time-varying attributes (Step 400) and defining static attributes (Step 402). [0053] Defining time-varying attributes (Step 400) can further include determining a lifestyle attribute (Step 404), determining activity attributes (Step 406), and determining mobility attributes (Step 408). [0054] The lifestyle attribute can be determined using an unsupervised learning model. The model identifies similarities in activity patterns among users. In certain embodiments it can be a clustering and dimension reduction model, Hidden Markov model, or an LDA and topic model. In some certain embodiments, the LDA and topic model comprises an Author topic models (ATM) where users are the authors, activities are words, periods of activity are documents, and lifestyles are topics. [0055] The mobility attributes represent the mobility patterns of the users based on their location data. [0056] Defining static attributes (Step 402) can further include determining accessibility attributes (step 410) and determining social demographic attributes (Step 412). [0057] FIG. 5 depicts the steps involved in determining predisposition for a particular behavior or experience/condition and tuning (Step 214). It involves, for each user 102, modeling time-varying attributes (Step 500), combining modeled time-varying attributes with static attributes (Step 502), assigning a quantified predisposition for a particular behavior or experience/condition (step 504); and adjusting parameters based on results resulting in a trained model [0058] In certain embodiments, modeling time-varying attributes (Step 500) involves embedding the lifestyle attribute into a continuous vector representation of a fixed dimension having learnable weigh parameters and tunable model hyperparameters (Step 508), concatenating the embedded lifestyle attribute and the attributes for the activity and mobility (Step 510), and transforming the concatenated embedded lifestyle attribute and time varying numerical attributes into a hidden representation having shared learnable weight and bias parameters and a tunable model hyperparameter (Step 512). [0059] In some such embodiments, transforming the concatenated embedded time varying categorical attributes and time varying numerical attributes into a hidden representation (step 512) comprises using a supervised machine learning model that models both spatial and temporal information from a user trajectory. In specific embodiments, the 4868-3590-9304, v.1 supervised machine learning model comprises a non-deep-learning regression or classification model such as a decision tree based model, a random forest based model, or a gradient boosting model. In other embodiments, the supervised machine learning model comprises a deep learning model. In still other embodiments, the supervised machine learning model comprises a deep neural network based model such as one or more of Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), Long Short Term Memory (LSTM) models, Radial Basis Function Networks (RBFN), and Transformer- based attention models. In one specific embodiment, the deep neural network based model comprises a Convoluted Long Short Term Memory (CLSTM) model with the time varying attributes as inputs [0060] In certain embodiments, combining modeled time-varying attributes with static attributes (Step 502) involves embedding categorical static attributes into a continuous vector representation of a fixed dimension having learnable weigh parameters and tunable model hyperparameters (Step 514) and concatenating the embedded embedding categorical static attributes, numerical static attributes, and a hidden representation of the modeled time- varying attributes (Step 516). [0061] Assigning a quantified predisposition (Step 504) involves making a probabilistic determination for particular behavior or experience/condition. The predisposition for a particular behavior or experience/condition can be any number of behaviors, experiences, or conditions. For example, the predisposition for a particular behavior or experience/condition is purchase intent, hospitalization risk, job participation, job change, travel intent, residential relocation intent, or healthcare risk. [0062] In certain embodiments, the predisposition for a particular behavior or experience/condition is associated with an opportunity window such as months, weeks, or days, or hours. [0063] Adjusting parameters (Step 506) allows the model to be tuned so that the assigned quantified predisposition (Step 504) reflects the desired behavior or experience/condition that is to be predicted. In certain embodiments, model is tuned via cross-validation. [0064] Referring back to FIG. 2, once the model has been trained (step 202), the trained model can then be used to predict a predisposition for a particular behavior or experience/condition for an individual user (Step 204). This step includes obtaining location data for the individual user (Step 216), inputting the location data into the trained model (step 4868-3590-9304, v.1 218), and receiving an assigned quantified predisposition for a particular behavior or experience/condition for the user from the trained model (step 220). [0065] In certain embodiments, the predisposition is then outputted (Step 222). In some such embodiments, outputting involves providing a graphical representation of the assigned quantified predisposition for a particular behavior or experience/condition for the particular user. In still other embodiments, outputting involves providing a web service API providing access to the assigned quantified predisposition for a particular behavior or experience/condition for the particular user. [0066] FIGs. 6-9 depict various examples of a graphical output that could be provided. FIG. 6 is an illustrative dashboard graphically showing user predisposition in a region. FIG. 7 is an illustrative dashboard where it has zoomed in to graphically show user predisposition in a more specific region. FIG. 8 is an illustrative dashboard graphically showing real-time predisposition modeling for a user. FIG. 9 is an illustrative dashboard graphically showing potential uses for such real-time in accordance with embodiments of the present invention. [0067] Any suitable and specifically configured electronic or computing device can be used to implement the functionality of the present invention including hardware or mobile device 108, server or cloud service 104 (including modules 112, 114, 116), business 106, and vendor 118 described herein. One illustrative example of such an electronic or computing device 1000 is depicted in FIG. 10. The computing device 1000 is merely an illustrative example of a suitable computing environment and in no way limits the scope of the present invention. A “computing device,” as represented by FIG. 10, can include a “workstation,” a “server,” a “laptop,” a “desktop,” a ”device”, a “smart device”, a “tablet”, a “smartphone”, an “ECR” or other specifically configured computing devices, as would be understood by those of skill in the art. Given that the computing device 1000 is depicted for illustrative purposes, embodiments of the present invention may utilize any number of computing devices 1000 in any number of different ways to implement a single embodiment of the present invention. Accordingly, embodiments of the present invention are not limited to a single computing device 1000, as would be appreciated by one with skill in the art, nor are they limited to a single type of implementation or configuration of the example computing device 1000. [0068] The computing device 1000 can include a bus or network 1010 that can be coupled to one or more of the following illustrative components, directly or indirectly: a memory 1012, one or more processors 1014, one or more presentation components 1016, input/output ports 1018, input/output components 1020, and a power supply 1024. 4868-3590-9304, v.1 [0069] One of skill in the art will appreciate that the bus 1010 can include one or more busses, such as an address bus, a data bus, networks, or any combination thereof. One of skill in the art additionally will appreciate that, depending on the intended applications and uses of a particular embodiment, multiple of these components can be implemented by a single device. Similarly, in some instances, a single component can be implemented by multiple devices. As such, FIG. 10 is merely illustrative of an exemplary computing device that can be used to implement one or more embodiments of the present invention, and in no way limits the invention. [0070] The computing device 1000 can include or interact with a variety of computer- readable media. For example, computer-readable media can include Random Access Memory (RAM); Read Only Memory (ROM); Electronically Erasable Programmable Read Only Memory (EEPROM); flash memory or other memory technologies; CDROM, digital versatile disks (DVD), Solid State Drive(SSD), cloud, or other optical or holographic media; magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices that can be used to encode information and can be accessed by the computing device 1000. [0071] The memory 1012 can include computer-storage media in the form of volatile and/or nonvolatile memory. The memory 1012 may be removable, non-removable, or any combination thereof. Exemplary hardware devices are devices such as hard drives, solid-state memory, optical-disc drives, and the like. The computing device 1000 can include one or more processors that read data from components such as the memory 1012, the various I/O components 1020, etc. Presentation component(s) 1016 present data indications to a user or other device. Exemplary presentation components include a display device, speaker, printing component, vibrating component, etc. [0072] The I/O ports 1018 can enable the computing device 1000 to be logically coupled to other devices, such as I/O components 1020 using serial, parallel, or network, and/or wireless communication protocols. Some of the I/O components 1020 can be built into the computing device 1000. Examples of such I/O components 1020 include a microphone, joystick, recording device, gamepad, satellite dish, scanner, printer, wireless device, networking device, and the like. Illustrative Example: Purchase Intent [0073] In the following example, the assigned quantified predisposition is purchase intent. 4868-3590-9304, v.1 Problem [0074] Identifying customers' intent to shop from their geolocation data can enable businesses to target specific areas, tailor promotions, and enhance the overall shopping experience. By analyzing patterns and trends in customer behavior, businesses can proactively reach out to potential customers and offer relevant products or services, ultimately increasing sales and customer engagement. While many businesses rely on traditional marketing and demographic data, integrating geolocation data provides valuable insights into customer preferences, interests, and behaviors specific to their geographic location. This information can be used to develop location-based marketing campaigns, recommend nearby stores or services, and optimize inventory management based on demand in different areas. As the retail industry evolves, the shift towards understanding customer intent from geolocation data becomes increasingly important. [0075] By leveraging predictive analytics and machine learning techniques, businesses can anticipate customer needs, optimize resource allocation, and deliver a personalized shopping experience tailored to individual customers and their geographic location. In summary, predicting customer intent to shop from geolocation data is an emerging field that offers numerous benefits to businesses. By utilizing this data effectively, companies can gain a competitive edge, improve customer satisfaction, and drive sales by delivering targeted marketing campaigns and personalized shopping experiences. [0076] This invention offers a solution to this problem. This technology operationalizes geolocation data collected from mobile phones in a privacy-preserving manner and leverages proprietary cloud computation to predict customer visits to QSRs with greater efficacy than current state of the art techniques. Proposed Solution [0077] Embodiments of the present invention relate to a proactive computer simulation of likely business visit when no user demographics data or data on user prior behavior such as purchase spending or response to prior promotion is available. The invention is an end-to- end process of creating a proxy for personal consumption factors to make this prediction, accomplished by: (a) geolocation data collection from cellphones; (b) organizing and controlling this data; (c) computational analysis of the data in a privacy-preserving, cloud- based proprietary server; and (d) timely communication of the prediction to users and businesses through user applications and APIs. 4868-3590-9304, v.1 Framework [0078] The presented framework is aimed at extracting proxies for user taste and consumption preference from location data to simulate user choice for businesses. We now introduce notation and state our research question. [0079] Definition 1 (Trajectory) A trajectory ^_^ of an individual ^ is defined as a temporally ordered set of tuples is a location where ^_^ ^{^} and ^_^ ^{^} are the coordinates of the geographic location (Coordinates usually correspond to latitude and longitude), and ^_^ ^{^} is the corresponding time stamp. [0080] Definition 2 (Activity-Trajectory) An activity trajectory ^_^ of an individual ^ is defined as mapping ^_^ to activities that exhibit a pattern of behavior, consumption, leisure. ^_^is a temporally ordered set of tuples ^_^ = ^{^ ^ ^ ^} ^_^ = ^^_^, ^_^^ where ^_^ = ^^^^^_^ ^{^}^ ^_^ ^{^} ∈ ^_^ is an activity by the individual inferred from a location closest to ^_^ ^{^} and ^_^ ^{^}, and ^_^ ^{^} is a coarser timestamp (For instance, tⁱ _j = 9:33 AM is coarsened and represented as 9 - 11 AM) of ^_^ ^{^}. Also, denote ^ as the universe of all temporal activities ^_^ ^{^} across ^_^. [0081] Lifestyle, one of the proxies, captures the behavior of a consumer on a day-day basis. [0082] Definition 3 (Lifestyle) A lifestyle ^_^ of an individual ^ is defined as a set of activities and their corresponding timestamps ^_^ = ∈ ^, |^_^| = that globally represent an individual’s day to day temporal activities across ^_^. [0083] Next, we illustrate the transformation of consumer trajectories (^_^) to activity trajectories (^_^) (Locations and Trajectories). We detail the identification of lifestyles (^_^) from consumer location trajectories in Section 2.2.1. In Section 2.3, we discuss our methodology to simulate a business visit from ^_^, ^_^ and other proxies of user consumption data capturing consumer mobility, accessibility and social demographics. 4868-3590-9304, v.1 Table 1 Activity groups Locations to Activity trajectories [0084] Mapping consumer locations to points of interest (such as place types - restaurants, groceries, or business types - Walmart, Five guys, etc.) opens up a new realm of possibilities to understand and simulate consumers’ social and behavioral determinants of consumptions, from both macro and micro perspectives. For instance, macro movement and temporal patterns across different competing brands inferred from such mapping were used to decide the placement of a new franchise. Further, micro, day to day consumer-specific patterns such as the number of visits to place types or time spent at various business types can predict consumer’s next likely location. [0085] We design our mapping to capture consumer activities grounded in sociologist’s definition of lifestyle - an activity that exhibits a pattern of behavior, consumption or leisure. To achieve this, we map locations to place types (refer to second column of Table 1) leveraging publicly available resources, and use home, work definitions. Next, we group place types with similar semantics ( first column in Table 1) to form activity groups that represent consumer consumption (restaurant, unhealthy activities), leisure (recreation, personal care, hotel, home, fitness), shopping (necessity shopping, leisure shopping) and commute (public transport, own transport) behavior. These 15 types of activity groups form the universe of all activities ^_^ ^{^} [0086] Further, to abstract away variations of the exact time in day to day activities, a coarser timestamp of ^_^ ^{^} (granular timestamp associated with each consumer location), ^_^ ^{^} is associated with each activity : 12 - 2AM, 3-5AM, 5-7AM, 7-9AM, 9-11AM, 11-2PM, 2- 4868-3590-9304, v.1 5PM, 5-7PM, 7-9PM, 9-12 PM. The resulting tuples of ^_^ ^{^} = across consumer trajectories form the universe (^ as defined in Def. 2) of all temporal activities ^_^ ^{^}. A more detailed explanation of our mapping from location (^_^) to activity trajectories (^_^) is presented below. Proxies from Location and Activity Trajectories Lifestyle as determinants of consumption [0087] Lifestyles, measured as work, life, travel, physical activities, dietary, leisure, healthcare, consumptions, etc. can demonstrate strong relationship with individuals’ economic preference and future consumption behaviors. [0088] Automatic discovery and simulation of lifestyles from location data is a non- trivial problem given the massive scale and high dimensionality. Besides, the differences in an individual’s activities across days, and the differences from other individuals’ activities add further complexity. We take an unsupervised topic modeling approach that has shown potential for uncovering complex temporal and behavioral patterns to identify work, home, consumption routines on smaller location data sets. Specifically, we leverage concepts of probabilistic Author Topic Models (ATM), designed for text documents to model day-to-day activities of a consumer. A probabilistic graphical model of this model using plate notation can be seen in FIG. 11. Leveraging our granular location data, we extend this line of literature by incorporating an extensive set of fifteen types of activities that represent patterns of behavior, consumption, and leisure. [0089] Author Topic Models: LDA is a probabilistic, unsupervised learning model of a collection of bags and of hidden discrete variables called topics. For text modeling, we may view each document as a mixture of various topics, where each topic is characterized as a distribution over words. ATM subsumes LDA and assumes authors of documents represent a multinomial distribution over topics where each topic is a probability distribution over words. A document with multiple authors has a distribution over topics that is a mixture of the distributions associated with the authors. FIG. 11 displays the graphical model of ATM. When generating a document, an author is chosen at random for each individual word in the document. This author picks a topic from their multinomial distribution over topics and then samples a word from the multinomial distribution over words associated with that topic. We repeat the process for all words in the document. Formally, the probability of a word 4868-3590-9304, v.1 !_" assuming # topics, $ authors, ^ documents and ^ unique words is: % &!_"' = ∑_)*^:, &-_" = .'%&-_" = .' where -_"is a latent variable indicating the topic from which the ^^"/ word was drawn. The aim of ATM inference is to determine the word distribution %^&- = .^' = 0₁ ^&)' for each topic . and the distribution of topics for authors %^&^ = .^' = 2₎ ³ for each author ^. %^&2^' is a Dirichlet(4) and P (0) is a Dirichlet(5), where and 4and 5 are hyper-parameters. We use the Gibbs approximation to estimate these as where and >^& ₃ ^' ) are the number of times word ! and author ^ have been assigned to topic . respectively. Similarly, = ^∑ _^:; >₎ ¹ ∶, >_. ³ = ^∑ _^:, >₎ ³ are the word-topic and author-topic sum respectively. A more detailed primer on ATM and Gibbs approximation is presented below. Next, we detail our ATM based lifestyle identification from consumer activity trajectories (^_^). [0090] Activity trajectories to Lifestyles: To identify lifestyles, we make an analogy between text documents and day-to-day activities, authors, and consumers. We view each activity ^_^ ^{^} in ^_^, the mapped activity trajectory as a word !. We represent each day’s activities of a consumer (author) as a bag of words - document ^. We view multiple days of a consumer ^ as unique documents of an author a. Based on these, we estimate the two ATM model parameters 0 ^{@A^ B C} ) , 2₎ ^{^} using Eq. 1 which represents the probability of activity for each topic ., and the probability of topics . for each consumer ^, respectively. Given these probability distributions, we can rank activities for each topic (lifestyle) discovered. We can also rank topics for consumers, resulting in the discovery of their primary lifestyle. [0091] We represent each lifestyle as the top Y activities ranked by their relevance (Sievert and Shirley 2014) - a convex combination of topic-specific probability of each activity (first term in Eq. 2) and lift (second term in Eq. 2, D_AB is the empirical distribution of activity ^_^). 4868-3590-9304, v.1 [0092] Next, we assign the most probable topic from the estimated author-topic distribution 2₎ ^{^} as the primary lifestyle of a consumer. Combining this with the top activities ranked by relevance, we can represent a consumer i’s lifestyle as ^_^ = ∈ ^. This completes identification of the individual’s lifestyle ^_^ from ^_^. We detail the choice of various hyperparameters for our ATM based lifestyle identification below. Next, we discuss the proposed sequential deep learner to simulate likelihood of future business visit. Business Visit Simulation [0093] To assess if a consumer has a visit to a business, we overlay day-to-day location trajectories of a consumer on publicly available location repositories of business facilities. [0094] To enrich our signals for simulation, we augment the identified lifestyles with other health care proxies extracted from location data. Proxies from Location and Activity Trajectories [0095] In Table 2, we describe different facets of consumer attributes we extract from location data and the proxies we used to indicate a consumer’s business visit outcome. [0096] Lifestyles: We identify consumer weekday and weekend lifestyles from their respective activity trajectories using the discussed ATM based technique. 4868-3590-9304, v.1 the daily groups ^_^ ^{^} as associate with health outcomes. This set of attributes capture daily consumer’s mobility patterns based on the locations visited in ^_^, such as the consumer’s frequency to, time spent at and distance traveled to a location. We also compute other richer mobility features, such as 4868-3590-9304, v.1 entropy and radius of gyration. All these are measured at a daily level and are time varying, numerical consumer attributes. [0099] Accessibility: To incorporate this facet of consumer characteristics, we leverage the transformed activity trajectory (^_^) and compute consumer accessibility - closest distance to various public facilities such as hospitals, parks, fitness centers, pharmacies, public transport and work from consumer’s home location. These attributes are static (time- invariant), numerical. [00100] Social Demographics : Based on the consumer’s home location from the transformed activity trajectories and publicly available census data, we also compute several block level social demographics. These are static and comprise of both categorical (for example employment poi, decoration of the work location of a consumer) and numerical attributes (population of consumer’s census block). Modeling Business Visit [00101] From Table 2, we see that we have multiple types of consumer attributes - time varying categorical (lifestyle, weekday/weekend) and numerical (work_freq, daily), static categorical (census_block_id) and numerical (commute_access). While the extensive nature of the attributes captures different facets of a consumer in relation to their business visit outcomes, this comes with several modeling challenges. First, one would need to jointly represent these unique types of attributes to model outcomes. Second, this representation would need to account for the possibility that interactions between these attributes lead to potentially better signals to predict health outcomes. For instance, temporal correlations between different day to day activity attributes, rather than day-level trends in each attribute could lead to a better predictive model. Third, given that these attributes capture multiple facets of a consumer attribute, naively concatenating all the representations may lead to a sub-optimal predictive model. For example, a naive way to represent all the attributes as a time series could be to concatenate the static categorical/numerical features to each timestamp, resulting in a highly parametric, potentially overfitted model. [00102] We address these issues by separately learning representations of time-varying and static features that account for interactions (temporal and static) among different types of attributes. Next, we combine these, allowing for interactions among the two representations, to learn a final joint representation of all the attributes. 4868-3590-9304, v.1 [00103] To achieve this, we represent time-varying attributes by a Context-LSTM (CLSTM) cell, a modification of the traditional LSTM cell, widely used for word translation and time series modeling. CLSTM, as the name reads, allows for incorporating both time- varying and static contextual features to a time series. In the original application, the time- varying contextual features considered were the latent topics of the words which were jointly represented with words (each word of the time series is concatenated with an embedding of the topic) to predict the next likely word in a sentence. Extending this to our setting, we note that lifestyles (Lifestyle attribute in Table 2), are latent topics learned from different activities. Hence, we view these as contexts to the different activity related time-varying attributes (Activity in Table 2). Further, we observe that viewing lifestyles as a context for the other time-varying attributes (Mobility in Table 2) leads to a better predictive performance empirically. Next, we concatenate these representations for a given time period with embeddings of static categorical and numerical attributes (Social Demographics and Accessibility) to jointly learn the representation of all consumer attributes that is predictive of the consumer health outcome. Such concatenation of multiple views of consumer attributes to form a unified representation are widely studied in multi-modal learning. [00104] An overview of the architecture diagram of the proposed sequential deep learning model is presented in FIG. 12. The upper left box in the figure illustrates the modeling of temporal attributes of a consumer at a day-level with a CLSTM cell (multiple days as a CLSTM layer) where the lifestyles are viewed as contexts for the activity and mobility attributes. The lower left box shows the representations of static consumer attributes which are later concatenated with the temporal representations to predict consumer health outcome. Next, we formally detail transformations performed by various layers in our proposed learner. Proposed Learner [00105] Let M_NO denote the time-varying numerical attribute tensor (num-ber of users number of days in the observation period number of time varying numerical features), M_NP the time-varying categorical attribute tensor (number of users number of weeks x number of time varying categorical attributes), matrices M_QO and M_QP (number of users x number of categorical/numerical features) denote the static numerical, categorical consumer attributes respectively. To simplify the notation, in the following discussion, we will focus on a single consumer’s attributes denoted by ^_NP , ^_NO , ^_QO and ^_QP and their transformation to probability of health outcome (health risk). 4868-3590-9304, v.1 [00106] Embedding: Embedding layers transforms one-hot encoded categorical attributes (^_NP, ^_QP) to a continuous vector representation of a fixed dimension. Formally, R_NP = ^_NP^_N ^S P (3) R_QP = ^_QP^_Q ^S P [00107] where ^_N ^S P - number of time varying categorical attributes x T_N ^S P , ^_Q ^S P - number of static categorical attributes x T_Q ^S P are the learnable weight parameters, T_N ^S P , T_Q ^S P are tunable model hyperparameters. Recall that in our setting, ^_NP comprises of weekday and weekend lifestyles (^_^), both represented by top ten relevant activities (^_^ ^{^}, universe of activities ^). Hence, a consumer’s weekday, weekend lifestyle can both be represented as a vectors of length |^|, meaning, we learn two weight matrices of dimensionality ^|^^|^ T_N ^S P W compute R_NP. A similar procedure is followed to transform the other static categorical attributes (employment_poi, census_block_id). [00108] CLSTM layer: a CLSTM layer as shown in FIG. 13 comprises of multiple CLSTM cells, each of which act on different days of the ^_NO , R_NP. Assume ^_N ^" O , R_N ^" P correspond to all numerical, embedded categorical time-varying consumer attributes on an arbitrary day (R_N ^" P is computed depending on whether the day was a weekday or weekend, since our lifestyles are derived for weekday/weekend rather than days), each CLSTM cell performs the following transformations (4) 4868-3590-9304, v.1 The above four equations detail modifications of the traditional LSTM cell where ^, U and G are the input, output and forget gates respectively to incorporate additional context R_N ^" P . Rearranging the terms, we note that this is equivalent to considering a composite input V^_N ^" O R_N ^" P W since (5) Each CLSTM cell transforms the concatenated input V^_N ^" O R_N ^" P W into ℎ_" (dimensions : number of consumers x T_N ^S ) with learnable shared weight and bias parameters &^_∗, Z' and tuneable hyperparameter T_N ^S . Hence, the resulting representations from the CLSTM layer are ^[ℎ^"\, ^ ∈ V1, ^W, where ^ is the number of days in our observation period. [00109] Concatenate: Concatenate layers do not contain any learnable parameters and are simply used to combine different intermediate representations. We perform two concatenations (refer Figure 2). The first one is discussed above where we construct a composite input V^_N ^" O R_N ^" P W into the CLSTM layer (as seen in FIG. 13). Next is the concatenation of the hidden temporal representation obtained from the CLSTM layer ([ℎ^"\), embedded static representation of days due to the Vℎ^NR_QP ^_QOW, the attributes. [00110] Shopping intent: We pass on the final representation into a fully connected dense layer, allowing for interactions between the temporal and static attributes and assign the quantified likelihood of shopper intent as E = ]&V^_N^_QP^_QO1WVℎ^NR_QP ^_QO Z_^W'^N (6) where ^_N , ^_QP , ^_QO , Z_^ are learnable parameters. For a given binary health outcome, to learn the various weights (denoted by ^_∗ in Equations 3, 5, 6), we minimize the binary cross- entropy loss between the observed shopping outcome (e.g. shopping_visit) and E, the vector 4868-3590-9304, v.1 of shopping intent from the above equation. The rest of the hyperparameters are tuned via cross-validation. These details are discussed below. Empirical Study [00111] GPS location Community urgent care collector that aggregates location data across hundreds of commonly used mobile applications ranging from news to weather, map navigation, and fitness. The location-data collection was performed through a GDPR- and CCPA-compliant framework. The data covers one-quarter of the U.S. population across the Android and iOS operating systems. Each row of the data corresponds to a location recorded for an individual. Each row contains information about: • Individual ID: an anonymized unique identifier of an individual using a mobile app, • Latitude, longitude, and timestamp of a location visited. • Speed at which the location was captured [00112] In total, we acquired location data of individuals from Boston across five months (September to January) in 2019. We only consider consumers who appear across all the four months and were tracked for more than ten days in each month. Further, we drop the consumers for whom we were unable to estimate a work and home location based on the heuristic discussed in Appendix B. Our final data set comprises of over 23,000 individuals. In Tables 3, 4 and Figure 4, we detail and provide the summary statistics for the different types of consumer attributes computed from location trajectories, their activity mappings, census block and public medical facilities data. We discuss them in detail next. Summary Statistics [00113] Location Trajectories: In Table 3 (Mobility row), we present summary statistics of the raw location per consumer, per day, for the four months of which around 18 are unique. The average speed at which these was 6.92 kmph. For the rest of the measures, we drop the locations that were captured at a speed of greater than 5 kmph and only consider stay locations - 4868-3590-9304, v.1 locations where consumers have spent at least five minutes. The average haversine distance between stay locations is 7.82 km and the average time spent at these locations is around 2.16 hours. Overall, the location data summary statistics indicate granular consumer observations across both cities. [00114] Activity Trajectories : Table 3 (Activity and Accessibility rows), Table 4 detail the summary statistics of the activity trajectories (Refer to Section 2.1 for the transformation of location to activity trajectories). From Table 3, we observe that out of the 15 pre-defined activity groups (Refer Table 1), home, work and public transport are the top three activity groups in terms of both avg. daily occurrences and time spent. When broken down by weekday and weekend (Table 4), as expected, we observe that work occurs less frequently (0.79) during the weekend compared to weekdays (4.27). Also, home occurs more frequently during weekends (5.57) compared to weekdays (8.97). To accommodate the differences in the top occurring activities, we separately learn weekday and weekend lifestyles for our empirical analysis. For weekdays, on average, we have 14.12 daily activities per consumer, translating to an average of 14 words per document based on our bag of words representation of a consumer’s activities for lifestyle identification (As discussed above). 4868-3590-9304, v.1 _{Feature grouping Name Me} ^Boston Std. Mi Ma an Dev. n x Weekend and weekday Lifestyle lifestyle lifestyles home freq 5.72 16.1 1 229 home dwell 5.87 3.71 0 24 health freq 1.21 8.24 0 189 health dwell 0.45 1.72 0 7.29 necessityshopping freq 1.62 8.21 0 241 necessityshopping dwell 0.61 1.55 0 7.72 publictransport freq 1.51 7.21 0 218 publictransport dwell 0.38 1.24 0 6.15 religious freq 0.78 4.41 0 167 religious dwell 0.26 1.07 0 5.53 work freq 3.72 8.18 0 213 work dwell 4.32 2.36 0 24 A^{ctivity hospital freq 0.13 2.20 0 164} h_{ospital dwell 0.04 0.43 0 24} personalcare freq 0.30 3.47 0 208 personalcare dwell 0.09 0.59 0 2.55 restaurant freq 0.78 4.22 0 145 restaurant dwell 0.27 0.83 0 3.92 unhealthyactivites freq 0.16 2.21 0 140 unhealthyactivities dwell 0.04 0.39 0 4.32 leisureshopping freq 0.26 2.45 0 149 leisureshopping dwell 0.10 0.59 0 4.91 hotel freq 0.32 2.68 0 122 hotel dwell 0.04 0.46 0 24 owntransport freq 0.24 2.89 0 167 owntransport dwell 0.12 0.63 0 24 n locations 31.2 40.4 3 1585 avg distance 7.82 6.70 0 126 avg location entropy 1.90 1.11 0 1 M_obility ^{avg time entropy 1.68 1.17 0 5.46} n unique locations 18.1 15.7 1 573 0.0 avg time spent 2.16 3.64 8 24 125. avg rog 6.11 4.28 0 1 avg speed 6.92 10.57 0 129 0.0 hospital access 1.63 0.84 2 2.62 0.0 park access 0.40 0.29 4 2.67 0.0 A_ccessibility ^{tness access 0.55 0.37} 2 2.04 0.0 prescription access 0.45 0.28 2 2.05 0.0 commute access 0.15 0.13 2 3.05 work access 1.90 3.34 0 39.4 0.4 employment percent 0.82 0.09 4 1 health ins percent 0.99 0.04 0 1 Social population 1145 561 3 4696 5832 86 2500 Demographics household income 1 31951 54 00 10. median age 37.4 9.30 8 79.9 gross rent 240 241 0 1384 Table 3 Summary Statistics of Consumer Attributes 4868-3590-9304, v.1 On weekends, this number remains similar. The average number of unique daily activities is 10.16 and 9. [00115] I e activity trajectories d FIGs 14a and 14b, a ce the correspondin work appear in 2 - 5 P.M., s likely to stay at home betw . 14b) we observe that riod. Also, other activiti g 9 - 11 P.M.) on wee ivity group unhealthyact compared to 12 - 5 A.M. [00116] he Haversine distance betw ude and longitude of hics) details the summary [00117] nded to be construed as plary”, “example”, a ance, or illustration” rred or advantageou the terms “about”, “ge may existing in the upper h as variations in example, the terms “about less, or minus 10 percent or , and “approximat in the relevant field to be in mplete or nearly complete extend or degree of an action, characteristic, property, state, structure, item, or result, as would be appreciated by one of skill in the art. For example, an object that is “substantially” circular would mean that the object is either completely a circle to 4868-3590-9304, v.1 mathematically determinable limits, or nearly a circle as would be recognized or understood by one of skill in the art. The exact allowable degree of deviation from absolute completeness may in some instances depend on the specific context. However, in general, the nearness of completion will be so as to have the same overall result as if absolute and total completion were achieved or obtained. The use of “substantially” is equally applicable when utilized in a negative connotation to refer to the complete or near complete lack of an action, characteristic, property, state, structure, item, or result, as would be appreciated by one of skill in the art. [00118] Numerous modifications and alternative embodiments of the present invention will be apparent to those skilled in the art in view of the foregoing description. Accordingly, this description is to be construed as illustrative only and is for the purpose of teaching those skilled in the art the best mode for carrying out the present invention. Details of the structure may vary substantially without departing from the spirit of the present invention, and exclusive use of all modifications that come within the scope of the appended claims is reserved. Within this specification embodiments have been described in a way which enables a clear and concise specification to be written, but it is intended and will be appreciated that embodiments may be variously combined or separated without parting from the invention. It is intended that the present invention be limited only to the extent required by the appended claims and the applicable rules of law. [00119] It is also to be understood that the following claims are to cover all generic and specific features of the invention described herein, and all statements of the scope of the invention which, as a matter of language, might be said to fall therebetween. 4868-3590-9304, v.1 mathematically determinable limits, or nearly a circle as would be recognized or understood by one of skill in the art. The exact allowable degree of deviation from absolute completeness may in some instances depend on the specific context. However, in general, the nearness of completion will be so as to have the same overall result as if absolute and total completion were achieved or obtained. The use of “substantially” is equally applicable when utilized in a negative connotation to refer to the complete or near complete lack of an action, characteristic, property, state, structure, item, or result, as would be appreciated by one of skill in the art.

[00118] Numerous modifications and alternative embodiments of the present invention will be apparent to those skilled in the art in view of the foregoing description. Accordingly, this description is to be construed as illustrative only and is for the purpose of teaching those skilled in the art the best mode for carrying out the present invention. Details of the structure may vary substantially without departing from the spirit of the present invention, and exclusive use of all modifications that come within the scope of the appended claims is reserved. Within this specification embodiments have been described in a way which enables a clear and concise specification to be written, but it is intended and will be appreciated that embodiments may be variously combined or separated without parting from the invention. It is intended that the present invention be limited only to the extent required by the appended claims and the applicable rules of law.

[00119] It is also to be understood that the following claims are to cover all generic and specific features of the invention described herein, and all statements of the scope of the invention which, as a matter of language, might be said to fall therebetween.

Claims

CLAIMS What is claimed is: 1. A method of modeling a user predisposition based on location data, the method comprising: I. training at least one model to determine a user’s predisposition for a particular behavior or experience/condition, the training comprising: A) obtaining location data for multiple users; B) formatting location data into one or more trajectories for each user of multiple users; C) inferring activity data from location data; D) translating activity data and location data into time-varying and static behavior attributes for each user of the multiple users; and E) determining predisposition for a particular behavior or experience/condition comprising: for each user: modeling time-varying attributes; combining modeled time-varying attributes with static attributes; assigning a quantified predisposition for a particular behavior or experience/condition; and adjusting parameters based on results resulting in a trained model; and II. predicting predisposition for a particular behavior or experience/condition for an individual user comprising: A) obtaining location data for the individual user; B) inputting the location data into the at least one trained model produced by the training of at least one model; and C) receiving an assigned quantified predisposition for a particular behavior or experience/condition for the individual user from the trained model.

2. The method claim 1, wherein the predisposition for a particular behavior or experience/condition is purchase intent. 4868-3590-9304, v.1

3. The method of claim 1, wherein the predisposition for a particular behavior or experience/condition is hospitalization risk.

4. The method of claim 1, wherein the predisposition for a particular behavior or experience/condition is job participation.

5. The method of claim 1, wherein the predisposition for a particular behavior or experience/condition is job change.

6. The method of claim 1, wherein the predisposition for a particular behavior or experience/condition is travel intent.

7. The method of claim 1, wherein the predisposition for a particular behavior or experience/condition is residential relocation intent.

8. The method of claim 1, wherein the predisposition for a particular behavior or experience/condition is healthcare risk.

9. The method of claim 1, wherein the predisposition for a particular behavior or experience/condition is associated with an opportunity window.

10. The method of claim 1, wherein the location data is geospatial data for a user hardware device over a period of time.

11. The method of claim 1, wherein the location data for a user device is provided by a location provider/vendor.

12. The method of claim 1, wherein the location data for a user is identified without using one or more of: personal identifying information (PII), demographic information, and socioeconomic information about the user. 4868-3590-9304, v.1

13. The method of claim 1, wherein inferring activity data from location data further comprises: i) mapping location data to place types; ii) grouping place types into activity groups based on a function of the place type; and iii) transforming the one or more trajectories for the user into one or more activity- trajectories for the user using the activity groups.

14. The method of claim 13, wherein the activity groups comprise one or more selected from the group comprising: hospital, health, necessity shopping, fitness, public transport, own transport, religious, recreation, travel, personal care, leisure shopping, unhealthy activities, restaurant, home, and work.

15. The method of claim 1, wherein translating activity data and location data into behavior attributes for each user of the multiple users comprises: defining time-varying attributes; and defining static attributes.

16. The method of claim 15, wherein defining time-varying attributes comprises: determining a lifestyle attribute; determining activity attributes; and determining mobility attributes.

17. The method of claim 16, wherein determining a lifestyle attribute comprises using an unsupervised learning model that can identify similarities in activity patterns among users.

18. The method of claim 17, wherein the unsupervised learning model comprises a clustering and dimension reduction model.

19. The method of claim 17, wherein the unsupervised learning model comprises a Hidden Markov model.

20. The method of claim 17, wherein the unsupervised learning model comprises an LDA and topic model. 4868-3590-9304, v.1

21. The method of claim 20 wherein LDA and topic model comprises an Author topic models (ATM) where users are authors, activities are words, periods of activity are documents, and lifestyles are topics.

22. The method of claim 15 wherein defining static attributes comprises: determining accessibility attributes; and determining social demographic attributes.

23. The method of claim 16, wherein defining time-varying attributes comprises: embedding the lifestyle attribute into a continuous vector representation of a fixed dimension having learnable weigh parameters and tunable model hyperparameters; concatenating the embedded lifestyle attribute and the attributes for the activity and mobility; and transforming the concatenated embedded lifestyle attribute and time varying numerical attributes into a hidden representation having shared learnable weight and bias parameters and a tunable model hyperparameter.

24. The method of claim 23, wherein transforming the concatenated embedded lifestyle attribute and time varying numerical attributes into a hidden representation comprises using a supervised machine learning model that models both spatial and temporal information from a user trajectory.

25. The method of claim 24, wherein the supervised machine learning model comprises a non-deep-learning regression or classification model.

26. The method of claim 25, wherein the non-deep-learning regression or classification model comprises a decision tree based model, a random forest based model, or a gradient boosting model.

27. The method of claim 24, wherein the supervised machine learning model comprises a deep learning model. 4868-3590-9304, v.1

28. The method of claim 24, wherein the supervised machine learning model comprises a deep neural network based model.

29. The method of claim 28, wherein the deep neural network based model comprises at least one of: Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), Long Short Term Memory (LSTM) models, Radial Basis Function Networks (RBFN), Transformer-based attention models.

30. The method of claim 29 wherein deep neural network based model comprises a a Convoluted Long Short Term Memory (CLSTM) model with the time varying attributes as inputs.

31. The method of claim 15, wherein combining modeled time-varying attributes with static attributes comprises: embedding categorical static attributes into a continuous vector representation of a fixed dimension having learnable weigh parameters and tunable model hyperparameters; and concatenating the embedded categorical static attributes, numerical static attributes, and a hidden representation of the modeled time-varying attributes.

32. The method of claim 1, wherein the model is tuned via cross-validation.

33. The method of claim 1, further comprising outputting the assigned quantified predisposition for a particular behavior or experience/condition for the particular user.

34. The method of claim 33, wherein the outputting comprises providing a graphical representation of the assigned quantified predisposition for a particular behavior or experience/condition for the particular user.

35. The method of claim 33, wherein the outputting comprises providing a web service API providing access to the assigned quantified predisposition for a particular behavior or experience/condition for the particular user.

36. A system for modeling user predisposition based on location data, the system comprising: 4868-3590-9304, v.1 a data collection module configured to obtain location data for multiple users; a model training module configured to train a model to determine a user's predisposition for a particular behavior using the location data; and a prediction module configured to predict predisposition for a particular behavior for an individual user using the trained model and location data for the individual user.

37. The system of claim 36, wherein the model training module is further configured to infer activity data from location data, translate activity data and location data into behavior attributes, and determine predisposition for a particular behavior for each user.

38. The system of claim 36, wherein the prediction module is further configured to input the location data into the trained model and receive an assigned quantified predisposition for a particular behavior for the particular user from the trained model. 4868-3590-9304, v.1