HK1248895B - Risk characteristic screening and describing message generation method and device and electronic equipment - Google Patents

Description

Risk feature screening, description message generation method, device and electronic equipment

Technical Field

This specification relates to the field of computer technology, and in particular to risk feature screening, description message generation methods, devices, and electronic devices.

Background Art

With the rapid development of internet finance, the number of internet financial transactions is also growing rapidly. Among these large numbers of internet financial transactions, there is a risk that some individuals may engage in illegal activities such as money laundering. Therefore, personnel are required to identify suspicious transactions within the vast volume of transaction records, generate corresponding suspicious transaction descriptions, and report these to the relevant management departments. These suspicious transactions can also be considered risk events.

In the prior art, after receiving suspicious transaction data, staff usually manually write a message describing the suspicious transaction based on the data and a predefined message template, wherein the message length is limited.

Based on the existing technology, there is a need for a solution that can generate more reference descriptive messages for suspicious transactions based on message length constraints.

Summary of the Invention

The embodiments of this specification provide a method, device, and electronic device for risk feature screening and description message generation to solve the following technical problem: a solution is needed to generate a more reference-based description message for suspicious transactions based on message length constraints.

To solve the above technical problems, the embodiments of this specification are implemented as follows:

The present invention provides a method for screening risk characteristics, including:

Obtaining feature weights of each of the plurality of risk features, the feature weights being obtained based on a classification model trained using sample events or being predefined, the classification model being used to determine risk events;

At least some of the risk features are screened out according to the feature weights and predetermined conditions, where the predetermined conditions are used to constrain the length of messages generated according to the risk features.

The embodiment of this specification provides a description message generation method, including:

Get the event to be described;

Determine the risk characteristics of each screened risk;

generating a description message for the event to be described based on the screened risk features;

Among them, the screening out of each risk feature includes: obtaining the feature weights of multiple risk features respectively, and screening out the risk features according to the feature weights and predetermined conditions, the feature weights are obtained according to a classification model trained using sample events or are predefined, the classification model is used to determine risk events, and the predetermined conditions are used to constrain the length of the message generated according to the risk feature.

The embodiments of this specification provide a risk feature screening device, comprising:

an acquisition module, which acquires feature weights of a plurality of risk features, wherein the feature weights are obtained according to a classification model trained using sample events or are predefined, and the classification model is used to determine risk events;

The screening module screens out at least some of the risk features according to the feature weights and predetermined conditions, wherein the predetermined conditions are used to constrain the length of the message generated according to the risk features.

An embodiment of this specification provides a description message generation device, including:

Get the module and get the event to be described;

Determine the module and identify the risk characteristics of each screened out;

A generating module, generating a description message for the event to be described according to the screened risk features;

Among them, the screening out of each risk feature includes: obtaining the feature weights of multiple risk features respectively, and screening out the risk features according to the feature weights and predetermined conditions, the feature weights are obtained according to a classification model trained using sample events or are predefined, the classification model is used to determine risk events, and the predetermined conditions are used to constrain the length of the message generated according to the risk feature.

An embodiment of this specification provides an electronic device for screening risk characteristics, including:

at least one processor; and,

a memory communicatively connected to the at least one processor; wherein,

The memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to:

Obtaining feature weights of each of the plurality of risk features, the feature weights being obtained based on a classification model trained using sample events or being predefined, the classification model being used to determine risk events;

At least some of the risk features are screened out according to the feature weights and predetermined conditions, where the predetermined conditions are used to constrain the length of messages generated according to the risk features.

An electronic device for generating a description message provided in an embodiment of this specification includes:

at least one processor; and,

a memory communicatively connected to the at least one processor; wherein,

The memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to:

Get the event to be described;

Determine the risk characteristics of each screened risk;

generating a description message for the event to be described based on the screened risk features;

Among them, the screening out of each risk feature includes: obtaining the feature weights of multiple risk features respectively, and screening out the risk features according to the feature weights and predetermined conditions, the feature weights are obtained according to a classification model trained using sample events or are predefined, the classification model is used to determine risk events, and the predetermined conditions are used to constrain the length of the message generated according to the risk feature.

At least one of the above-mentioned technical solutions adopted in the embodiments of this specification can achieve the following beneficial effects: the classification model obtained by training can be used to determine the feature weights of each risk feature, and based on the feature weights and predetermined conditions for constraining the length of the message generated based on the risk features, a description message is generated for the event to be described, so that the generated description message is more useful for reference; wherein, the event to be described may be, for example, a suspicious transaction such as a suspected money laundering transaction.

BRIEF DESCRIPTION OF THE DRAWINGS

In order to more clearly illustrate the embodiments of this specification or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments recorded in this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

FIG1 is a schematic diagram of an overall architecture of the solution of this specification in a practical application scenario;

FIG2 is a schematic diagram of a process for screening risk characteristics according to an embodiment of the present disclosure;

FIG3 is a flow chart of a description message generation method provided in an embodiment of this specification;

FIG4 is a schematic diagram of a partial screenshot of a description message provided in an embodiment of this specification;

FIG5 is a schematic diagram of an automatic message algorithm provided in an embodiment of this specification;

FIG6 is a schematic diagram of a suspicious transaction identification process in an actual application scenario provided by an embodiment of this specification;

FIG7 is a schematic structural diagram of a risk feature screening device corresponding to FIG2 provided in an embodiment of this specification;

FIG8 is a schematic structural diagram of a description message generating device corresponding to FIG3 provided in an embodiment of this specification.

DETAILED DESCRIPTION

The embodiments of this specification provide a method, device, and electronic device for screening risk characteristics and generating description messages.

In order to help those skilled in the art better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments of this specification, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of this application.

To facilitate understanding, the idea of the solution in this specification is analyzed.

In the absence of message length constraints, the message describing the suspicious transaction can cover all information points, where each information point reflects data related to one of the risk characteristics of the suspicious transaction. For example, an information point is a sub-message generated based on a risk characteristic. The set of all risk characteristics is denoted as S.

However, when message length is constrained, the description message typically only covers a portion of the risk characteristic data for suspicious transactions, rather than all of it, otherwise the message length will exceed the limit. Therefore, to maximize the reference value of the generated description message, the risk characteristics must be screened to identify the subset of risk characteristics with the highest reference value. This subset of risk characteristics is denoted as the reference value S', measured by the area under the receiver operating characteristic curve (AUC) of the assumed classification model. An ideal goal is to select the S' that maximizes the corresponding AUC.

This ideal goal belongs to a combinatorial optimization problem. When the number of risk features is large, the amount of calculation is very large and not conducive to practical application. Based on this, the solution of this specification uses a greedy search strategy to approximate the solution to the combinatorial optimization problem and obtain a local optimal solution. This can reduce the amount of calculation and is more efficient.

The solution of this specification can be used to: screen risk features with relatively high reference value from a set of risk features to be screened; further, it can be used to use the screened risk features to generate descriptive messages for risk events such as suspicious transactions.

Figure 1 is a schematic diagram of the overall architecture involved in the solution of this specification in an actual application scenario. This overall architecture includes at least one device, and the device workflow primarily includes: determining multiple risk features to be screened, and screening out at least some of the risk features; and inputting an event to be described into a device for generating a description message. The device generates a description message based on the event to be described and the screened risk features. The at least one device may include a classification model for determining risk events.

Based on the above ideas and overall architecture, the solution of this manual is described in detail below.

The present invention provides a method for screening risk characteristics. As shown in FIG2 , the method may include the following steps:

S202: Obtain feature weights of the plurality of risk features, wherein the feature weights are obtained according to a classification model trained using sample events or are predefined, and the classification model is used to determine risk events.

In the embodiments of this specification, there are multiple sample events. For the same risk feature, the feature values of different sample events can be different. Generally, a classification model can be pre-trained using sample events, and then the classification model can be used to determine the feature weights corresponding to each risk feature.

For example, the feature weight can be obtained by calculating the classification accuracy metric of the risk feature corresponding to the classification model, where the classification accuracy metric is, for example, AUC, information entropy, or classification precision.

Of course, feature weights can also be predefined without relying on the classification model.

Feature weights reflect the importance of risk features. Generally, risk features with higher feature weights are preferred for event description. Furthermore, due to message length constraints (the aforementioned predefined conditions), feature weights may not be the sole basis for screening risk features. For example, factors such as the length of the sub-message corresponding to the risk feature can also be considered for screening.

Risk events can be suspicious transactions, such as suspected money laundering or transactions conducted by a suspected account hacker impersonating the account owner. Risk events can also be suspicious business operations other than transactions, such as illegal logins.

S204: Filter out at least some risk features according to the feature weights and predetermined conditions, where the predetermined conditions are used to constrain the length of messages generated according to the risk features.

By using the method of Figure 2, risk features with greater reference value can be screened out. Based on the method of Figure 2, the present specification also provides some specific implementation plans and extension plans of the method, which are described below.

In the embodiments of this specification, pre-defining feature weights is relatively easy to understand and can generally be done based on the experience of the operator. The following mainly describes another method for obtaining feature weights.

For step S202, the feature weight is obtained by using the classification model obtained by training with sample events, which may specifically include: obtaining the classification model by using sample events; performing respectively for the multiple risk features: obtaining data corresponding to the risk feature in the sample events; calculating the classification accuracy measurement index of the risk feature corresponding to the classification model based on the data corresponding to the risk feature; and obtaining the feature weight of the risk feature based on the classification accuracy measurement index.

In the embodiments of this specification, the classification accuracy metric for a classification model corresponding to a risk feature may specifically represent the accuracy of classifying a sample event using the data corresponding to the risk feature as input to the classification model. For example, the AUC is a classification accuracy metric. A higher AUC indicates a higher classification accuracy.

The classification model can be a random forest model or a logistic regression model. Taking the random forest model as an example, suppose the training sample D = (x, y), where x∈Rn ^*d is the model input data; y∈Rn*1 is the sample label, which indicates, for example, whether the sample event involves money laundering, that is, whether it is a suspected money laundering transaction; then, based on the training sample data x and the sample label y, a decision tree is constructed, and the random forest model is trained based on the constructed multiple decision trees.

In the embodiment of the present specification, corresponding sub-messages can be generated based on the risk characteristic data. The plurality of risk characteristics each have a corresponding number of sub-message characters, and the number of sub-message characters can be predetermined or estimated.

In this case, for step S204, filtering out at least part of the risk features based on the feature weights and predetermined conditions may specifically include: performing a first sorting of the multiple risk features based on the feature weights and the corresponding number of sub-report characters; filtering out at least part of the risk features based on the first sorting result, the number of sub-report characters, and the predetermined conditions.

For example, the sub-message character count is a predefined character count for a sub-message template defined for a risk characteristic. The sub-message template can include risk characteristics and corresponding description statements. A predefined correspondence between each risk characteristic and description statement can be established; for example, <Feature 1, Description Statement 1>, <Feature 2, Description Statement 2>, <Feature 3, Description Statement 3>. Generally, the sub-message is generated by substituting the specific values of the risk characteristics into the description statements. The default character count for the description statements is then the predefined character count.

Furthermore, the first sorting of the multiple risk features according to the feature weights and the corresponding number of sub-report characters may specifically include: determining a second sorting result obtained by performing a second sorting of the multiple risk features according to the feature weights; selecting at least part of the multiple risk features according to the second sorting result; and performing a first sorting of the selected risk features according to the feature weights and the corresponding number of sub-report characters.

In practical applications, when there are many risk features, the risk features can be sorted and/or pre-screened before formal screening. This helps reduce the processing resources consumed by screening.

For example, assuming that the risk features are sorted in a second order according to the feature weights from large to small, the risk features at the back of the second sorting result can be eliminated, and the risk features at the front can be retained.

It should be noted that pre-screening (performed based on the second sorting described above) is not a necessary step, and whether to perform it can be determined based on actual needs.

In an embodiment of the present specification, the first sorting of the multiple risk features based on the feature weights and the corresponding number of sub-report characters may specifically include: calculating the unit word count weight corresponding to the risk feature based on the feature weights and the corresponding sub-report characters of the risk feature; and performing a first sorting of the multiple risk features according to the unit word count weight.

The unit word weight can be expressed as the average contribution of each word in the sub-message to its corresponding feature weight. More intuitively, for example, the unit word weight can be equal to the feature weight divided by the number of words in the corresponding sub-message.

Of course, risk features can also be sorted and screened based on other indicators besides the unit word weight, such as the amount of information per unit word.

As mentioned earlier in the solution, we use a greedy search strategy to approximate the solution. Below, we first illustrate the approximate solution process, followed by analysis.

In the embodiment of this specification, screening out at least some risk features based on the first sorting result, the number of sub-report characters, and a predetermined condition may specifically include:

Based on the first sorting result, for each risk feature included in the first sorting result, traverse in descending order of unit word weight, and execute the following for the current risk feature:

Add the current risk feature to the setting set, and determine whether the sum of the number of sub-report characters corresponding to the risk features included in the setting set meets the predetermined conditions; if so, traverse to the next risk feature; otherwise, remove the current risk feature from the setting set, end the traversal process, and use the risk features included in the setting set as at least part of the screened risk features; wherein, the setting set is initially an empty set.

In actual applications, in the above judgment process, if the judgment result is negative, it is not necessary to end the traversal operation. For example, you can continue to try to select the subsequent risk features in sequence to add to the set set and then see whether the constraints are met.

In the embodiment of this specification, for step S206, traversing to the next risk feature may specifically include:

Determining a classification accuracy metric corresponding to the classification model for the setting set;

Determine whether the classification accuracy metric is not greater than the classification accuracy metric of the setting set corresponding to the classification model before adding the current risk feature; if so, remove the current risk feature from the setting set and traverse to the next risk feature; otherwise, traverse to the next risk feature.

To avoid confusion, the following example illustrates the configuration set before the current risk feature is added. For example, if nine risk features have already been added to the configuration set (assuming the configuration set at this point is referred to as the "current set"), and the 10th risk feature (i.e., the "current risk feature") is to be added, then the configuration set before the current risk feature is referred to as the "current set."

The above shows the process of using the greedy search strategy to obtain an approximate solution, which is analyzed below.

To achieve the above-mentioned ideal goal, it is necessary to exhaustively enumerate the risk feature subset S' to find the S' with the maximum AUC (an example of a classification accuracy metric) while satisfying the message length constraint.

The greedy search strategy avoids exhaustive searching. Based on the first ranking result, it sequentially selects the best risk feature among the remaining risk features (in the above example, the best refers to the one with the highest weight per word count) until the message length constraint is met. Furthermore, it is assumed that the corresponding AUC increases with each risk feature added, thus avoiding the need to calculate the AUC each time, saving processing resources and improving screening efficiency.

Of course, to be more precise, the AUC can also be calculated each time. The reason is that a newly added risk feature may also reduce the AUC; in this case, the risk feature can be removed.

For example, if there is a risk feature S ⁽ⁱ⁾ that is highly correlated with the obtained set S', or S(i) contains obvious noise, and this risk feature S(i) will cause the classification ability of the classification model to decrease or remain unchanged (that is, the classification accuracy metric decreases or remains unchanged), then S ⁽ⁱ⁾ can be eliminated from S'.

In the embodiments of this specification, based on the screening of risk features, a description message can be further generated for risk events such as suspected money laundering transactions waiting to be described, wherein whether it is a risk event can be determined by the above-mentioned classification model or based on manual experience.

For example, an event to be described is obtained, and at least some risk features are screened out, and sub-messages corresponding to the event to be described are generated. The sub-messages are assembled to obtain a description message for the event to be described. In addition, to improve efficiency, predefined sub-message templates can be used to generate sub-messages.

Based on the same idea, the embodiment of this specification also provides a flowchart describing a message generation method, as shown in FIG3 .

The process in Figure 3 may include the following steps:

S302: Obtain the event to be described.

S304: Determine the risk characteristics of each screened out.

In the embodiments of this specification, risk features may be pre-screened before the process is executed, or may be screened after the event to be described is obtained.

S306: Generate a description message for the event to be described based on the screened risk features;

Among them, the screening out of each risk feature includes: obtaining the feature weights of multiple risk features respectively, and screening out the risk features according to the feature weights and predetermined conditions, the feature weights are obtained according to a classification model trained using sample events or are predefined, the classification model is used to determine risk events, and the predetermined conditions are used to constrain the length of the message generated according to the risk feature.

In actual applications, the corresponding sub-messages can be generated while the risk features are being screened, or the sub-messages can be generated after all risk features have been screened.

The method shown in FIG3 is helpful for generating a more informative description message for the event to be described.

More intuitively, the embodiments of this specification also provide an example of the content structure of a description message generated for a suspicious transaction. The description message includes, for example, six parts, each corresponding to one or more risk characteristics:

First, an overview of the suspicious transactions;

Second, describe the process of discovering the suspicious transaction, such as time, location, etc.

Third, the opening of suspicious accounts, such as basic information on account opening documents;

Fourth, the overall circumstances of the suspicious transactions, such as the time period of the transactions, the number and amount of transactions involved, the source and destination of the funds, and the transaction process;

Fifth, suspicious point analysis, listing suspicious points one by one, such as account opening and closing information and other suspicious information in the transaction process;

Sixth, the message is judged, and by combining all data analysis and subjective judgment, a final label is given to the transaction, for example, a suspected money laundering transaction.

Figure 4 is a schematic diagram of a partial screenshot of a description message provided by an embodiment of this specification, showing part of the six parts mentioned above. The description message generated based on the embodiment of this specification can highlight the key points and will not exceed the message length limit.

In a practical application scenario, two types of descriptive messages can be generated for suspected money laundering transactions. One type is the descriptive messages described in the above embodiments, known as deterministic messages. These messages are typically derived directly from objective data and are not mixed with subjective analysis data. The other type is known as nondeterministic messages, which may be mixed with subjective analysis data. In this case, the above message length constraints apply to deterministic messages.

The embodiments of this specification provide a modeling solution for automatically generating a description message model based on suspected money laundering transactions. The solution may include the following steps:

Given a labeled training sample set D(X,Y), where X∈R ^n*d is the sample model input data and Y∈R ^n*1 is the sample label. The sample label can indicate whether the sample event is a money laundering transaction.

Let S be the set of multiple risk features of the training sample, where |S| = d. Given a classification model f(D) for D, we hope to find a deterministic message corresponding to at least some of the risk features through this classification model, denoted as M(S'), such that the length of M(S') is no greater than a given threshold λ-θ, that is, |M(S')| ≤ λ-θ, where λ is the total constraint length of the deterministic message and the uncertain message, θ is the constraint length of the uncertain message, and λ-θ is the constraint length of the deterministic message (that is, the predetermined message length constraint condition mentioned above). Each constraint length is usually pre-set based on the actual situation (for example, different reviewers, different environments, etc.).

The ideal goal is to select an optimal feature set so that the AUC result AUC(D ^, S',f) of the data set corresponding to S ^* under the classifier f(D(S*)) is maximized, that is, to solve the following combinatorial optimization problem:

S ^* =argmax _|S'| AUC(D,S',f);

s.t.:|M(S')|≤λ-θ;

Among them, the objective function AUC(D,S',f) represents the AUC of D under the classifier f(X) after each feature subset S' is selected according to a certain scheme.

Of course, according to the previous analysis, it is costly to achieve this ideal goal. Therefore, as a second best option, a greedy search strategy is used to approximate the solution. FIG5 is a schematic diagram of an automatic message algorithm provided in an embodiment of this specification, which reflects the approximate solution process.

In Figure 5, the feature weight inverted list represents the second sorting result, S' represents the aforementioned set of settings, and step 3 represents the aforementioned process of traversing and screening risk features. It should be noted that in Figure 5, risk features are screened while sub-messages are generated. Upon completion of risk feature screening, the sub-messages that constitute the deterministic message are obtained.

Furthermore, the embodiments of this specification also provide a schematic diagram of a suspicious transaction identification process in an actual application scenario, as shown in FIG6 .

The process in Figure 6 mainly includes: generating a descriptive message generation task based on suspicious rules, wherein the task is for suspected money laundering transactions; further, the solution of this specification can be used to automatically execute the task (that is, generate a descriptive message for suspected money laundering transactions); and then manually conduct an initial review and a manual review of the descriptive message.

Based on the same idea, the embodiments of this specification also provide corresponding devices, as shown in Figures 7 and 8.

FIG7 is a schematic diagram of the structure of a risk feature screening device corresponding to FIG2 provided in an embodiment of this specification, including:

An acquisition module 701 acquires feature weights of a plurality of risk features, wherein the feature weights are obtained based on a classification model trained using sample events or are predefined, and the classification model is used to determine risk events;

The screening module 702 screens out at least some risk features according to the feature weights and predetermined conditions, where the predetermined conditions are used to constrain the length of messages generated according to the risk features.

Optionally, the apparatus further includes a weight determination module 703;

The weight determination module 703 obtains the feature weight according to the classification model trained using the sample events, specifically including:

The weight determination module 703 obtains a classification model by training sample events;

Execute for each of the risk features:

Obtaining data corresponding to the risk feature in the sample event;

According to the data corresponding to the risk feature, the risk feature corresponding to the classification is calculated.

Metrics for the model’s classification accuracy;

According to the classification accuracy measurement index, the feature weight of the risk feature is obtained.

Optionally, the plurality of risk features each have a corresponding number of sub-report characters; the screening module 702 screens out at least some of the risk features based on the feature weights and predetermined conditions, specifically including:

The screening module 702 performs a first sorting of the plurality of risk features according to the feature weights and the corresponding number of sub-report characters;

At least some risk features are screened out based on the first sorting result, the number of characters in the sub-report, and predetermined conditions.

Optionally, the screening module 702 performs a first sorting on the plurality of risk features according to the feature weights and the corresponding number of sub-report characters, specifically including:

The screening module 702 determines a second ranking result obtained by performing a second ranking on the plurality of risk features according to the feature weights;

selecting at least some of the risk features according to the second ranking result;

The selected risk features are first sorted according to the feature weights and the corresponding number of sub-report characters.

Optionally, the screening module 702 performs a first sorting on the plurality of risk features according to the feature weights and the corresponding number of sub-report characters, specifically including:

The screening module 702 calculates the unit word count weight corresponding to the risk feature according to the feature weight corresponding to the risk feature and the number of words in the sub-report;

The plurality of risk features are first sorted according to the unit word count weight.

Optionally, the screening module 702 screens out at least some risk features based on the first sorting result, the number of sub-report characters, and a predetermined condition, specifically including:

The screening module 702 traverses the risk features included in the first sorting result in descending order of unit word weight based on the first sorting result, and performs the following operations on the current risk feature:

Add the current risk feature to the setting set, and determine whether the sum of the number of sub-report characters corresponding to the risk features included in the setting set meets the predetermined conditions; if so, traverse to the next risk feature; otherwise, remove the current risk feature from the setting set, end the traversal process, and use the risk features included in the setting set as at least part of the screened risk features; wherein, the setting set is initially an empty set.

Optionally, the screening module 702 traverses to the next risk feature, specifically including:

The screening module 702 determines that the setting set corresponds to a classification accuracy metric of the classification model;

Determine whether the classification accuracy metric is not greater than the classification accuracy metric of the setting set corresponding to the classification model before adding the current risk feature; if so, remove the current risk feature from the setting set and traverse to the next risk feature; otherwise, traverse to the next risk feature.

Optionally, the classification accuracy metric includes the area under the receiver operating characteristic curve (AUC).

Optionally, the device further comprises:

The message generation module 704 obtains the event to be described;

Generating sub-messages corresponding to the event to be described for each of the at least some risk features screened out,

A description message is generated for the event to be described according to each of the sub-messages.

Optionally, the event to be described is determined by the classification model to be a risk event, and the risk event is a suspected money laundering transaction.

FIG8 is a schematic structural diagram of a description message generating device corresponding to FIG3 provided in an embodiment of this specification, including:

Acquisition module 801, acquires the event to be described;

Determination module 802, determining each risk feature screened out;

A generating module 803 generates a description message for the event to be described based on the screened risk features;

Among them, the screening out of each risk feature includes: obtaining the feature weights of multiple risk features respectively, and screening out the risk features according to the feature weights and predetermined conditions, the feature weights are obtained according to a classification model trained using sample events or are predefined, the classification model is used to determine risk events, and the predetermined conditions are used to constrain the length of the message generated according to the risk feature.

Based on the same idea, an embodiment of this specification further provides an electronic device, including:

at least one processor; and,

a memory communicatively connected to the at least one processor; wherein,

The memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to:

Obtaining feature weights of each of the plurality of risk features, the feature weights being obtained based on a classification model trained using sample events or being predefined, the classification model being used to determine risk events;

At least some of the risk features are screened out according to the feature weights and predetermined conditions, where the predetermined conditions are used to constrain the length of messages generated according to the risk features.

Based on the same idea, the embodiments of this specification also provide another electronic device, including:

at least one processor; and,

a memory communicatively connected to the at least one processor; wherein,

The memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to:

Get the event to be described;

Determine the risk characteristics of each screened risk;

generating a description message for the event to be described based on the screened risk features;

Among them, the screening out of each risk feature includes: obtaining the feature weights of multiple risk features respectively, and screening out the risk features according to the feature weights and predetermined conditions, the feature weights are obtained according to a classification model trained using sample events or are predefined, the classification model is used to determine risk events, and the predetermined conditions are used to constrain the length of the message generated according to the risk feature.

Based on the same idea, an embodiment of this specification further provides a non-volatile computer storage medium storing computer-executable instructions, wherein the computer-executable instructions are configured as follows:

Obtaining feature weights of each of the plurality of risk features, the feature weights being obtained based on a classification model trained using sample events or being predefined, the classification model being used to determine risk events;

At least some of the risk features are screened out according to the feature weights and predetermined conditions, where the predetermined conditions are used to constrain the length of messages generated according to the risk features.

Based on the same idea, the embodiments of this specification also provide another non-volatile computer storage medium storing computer-executable instructions, wherein the computer-executable instructions are configured as follows:

Get the event to be described;

Determine the risk characteristics of each screened risk;

generating a description message for the event to be described based on the screened risk features;

Among them, the screening out of each risk feature includes: obtaining the feature weights of multiple risk features respectively, and screening out the risk features according to the feature weights and predetermined conditions, the feature weights are obtained according to a classification model trained using sample events or are predefined, the classification model is used to determine risk events, and the predetermined conditions are used to constrain the length of the message generated according to the risk feature.

The foregoing description of this specification describes specific embodiments. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be performed in an order different from that described in the embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require the specific order shown or the sequential order to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

The various embodiments in this specification are described in a progressive manner. Similar portions between the various embodiments can be referenced to each other, and each embodiment focuses on the differences between the other embodiments. In particular, the device, electronic device, and non-volatile computer storage medium embodiments are generally similar to the method embodiments, so their descriptions are relatively simplified. For relevant details, refer to the descriptions of the method embodiments.

The apparatus, electronic device, and non-volatile computer storage medium provided in the embodiments of this specification correspond to the method. Therefore, the apparatus, electronic device, and non-volatile computer storage medium also have similar beneficial technical effects as the corresponding method. Since the beneficial technical effects of the method have been described in detail above, the beneficial technical effects of the corresponding apparatus, electronic device, and non-volatile computer storage medium will not be repeated here.

In the 1990s, technological improvements could be clearly distinguished as either hardware improvements (for example, improvements to circuit structures like diodes, transistors, and switches) or software improvements (improvements to process flows). However, with the advancement of technology, many process flow improvements today can now be considered direct improvements to hardware circuit structures. Designers almost always create the corresponding hardware circuit structure by programming the improved process flow into the hardware circuit. Therefore, it cannot be said that a process flow improvement cannot be implemented using hardware modules. For example, a programmable logic device (PLD), such as a field programmable gate array (FPGA), is an integrated circuit whose logical function is determined by user programming. Designers can "integrate" a digital system on a PLD through their own programming, without having to hire a chip manufacturer to design and manufacture a dedicated integrated circuit chip. Moreover, nowadays, instead of manually fabricating integrated circuit chips, this programming is mostly done using "logic compiler" software. This is similar to the software compiler used when developing programs. Before compilation, the original code must also be written in a specific programming language, called a hardware description language (HDL). There is not just one HDL, but many, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), Confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), Lava, Lola, MyHDL, PALASM, RHDL (Ruby Hardware Description Language), etc. The most commonly used ones are VHDL (Very-High-Speed Integrated Circuit Hardware Description Language) and Verilog. Those skilled in the art will also understand that by simply programming the method flow in one of these hardware description languages and then programming it into an integrated circuit, a hardware circuit that implements the logic method flow can be easily obtained.

The controller can be implemented in any suitable manner. For example, the controller can take the form of a microprocessor or processor and a computer-readable medium storing computer-readable program code (e.g., software or firmware) executable by the (micro)processor, logic gates, switches, application-specific integrated circuits (ASICs), programmable logic controllers, and embedded microcontrollers. Examples of controllers include, but are not limited to, the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicone Labs C8051F320. The memory controller can also be implemented as part of the control logic of the memory. Those skilled in the art will also know that in addition to implementing the controller in a purely computer-readable program code format, the controller can be implemented in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers by logically programming the method steps. Therefore, such a controller can be considered a hardware component, and the devices included therein for implementing various functions can also be considered as structures within the hardware component. Or even, the devices for implementing various functions can be considered as both software modules that implement the method and structures within the hardware component.

The systems, devices, modules, or units described in the above embodiments may be implemented by computer chips or entities, or by products having certain functions. A typical implementation device is a computer. Specifically, the computer may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smartphone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

For the convenience of description, the above devices are described as being divided into various units according to their functions. Of course, when implementing one or more embodiments of this specification, the functions of each unit can be implemented in the same or multiple software and/or hardware.

Those skilled in the art will appreciate that the embodiments of this specification may be provided as methods, systems, or computer program products. Therefore, the embodiments of this specification may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the embodiments of this specification may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

This specification is described with reference to the flowcharts and/or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of this specification. It should be understood that each process and/or box in the flowchart and/or block diagram, as well as the combination of the processes and/or boxes in the flowchart and/or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device produce a device for implementing the functions specified in one or more processes in the flowchart and/or one or more boxes in the block diagram.

These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a product including an instruction device that implements the functions specified in one or more processes in the flowchart and/or one or more boxes in the block diagram.

These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one or more processes in the flowchart and/or one or more boxes in the block diagram.

In a typical configuration, a computing device includes one or more processors (CPUs), input/output interfaces, network interfaces, and memory.

Memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and/or non-volatile memory in the form of read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.

Computer-readable media includes permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. The information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transitory computer-readable media (transitory media), such as modulated data signals and carrier waves.

It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or apparatus that includes a series of elements includes not only those elements but also other elements not explicitly listed, or includes elements inherent to such process, method, commodity, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity, or apparatus that includes the element.

This specification may be described in the general context of computer-executable instructions, such as program modules, executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, and the like that perform specific tasks or implement specific abstract data types. The specification may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media, including storage devices.

The various embodiments in this specification are described in a progressive manner. Similar parts between the various embodiments can be referred to in conjunction with each other. Each embodiment focuses on the differences between the other embodiments. In particular, the system embodiments are generally similar to the method embodiments, so the description is relatively simple. For relevant parts, refer to the description of the method embodiments.

The foregoing is merely an embodiment of the present invention and is not intended to limit the present application. For those skilled in the art, various modifications and variations may be made to the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included within the scope of the claims of the present application.

Claims (22)

1. A risk characteristic screening method, comprising: Obtain feature weights for multiple risk features. The feature weights are obtained based on a classification model trained using sample events or are predefined. The classification model is used to determine risk events. Based on the feature weights and predetermined conditions, at least some risk features are selected, wherein the predetermined conditions are used to constrain the length of the message generated based on the risk features; The plurality of risk features each have a corresponding number of sub-report characters; the step of selecting at least a subset of risk features based on the feature weights and predetermined conditions specifically includes: The multiple risk features are sorted in a first order based on the feature weights and the corresponding number of words in the sub-report; Based on the first sorting result, the number of words in the sub-report, and predetermined conditions, at least some risk characteristics are selected. 2. The method as described in claim 1, wherein the feature weights are obtained based on a classification model trained using sample events, specifically includes: A classification model is obtained by training sample events; Perform the following actions for each of the aforementioned risk characteristics: Obtain the data corresponding to this risk characteristic from the sample events; Based on the data corresponding to the risk feature, calculate the classification accuracy metric of the classification model corresponding to the risk feature; Based on the classification accuracy metric, the feature weight of this risk characteristic is obtained. 3. The method as described in claim 1, wherein the first sorting of the plurality of risk features according to the feature weights and the corresponding sub-report word counts specifically includes: The second ranking result is obtained by ranking the multiple risk features according to their weights. Based on the second sorting result, at least some of the risk features among the plurality of risk features are selected; The selected risk features are sorted in a first order based on the feature weights and the corresponding number of words in the sub-report. 4. The method as described in claim 1, wherein the first sorting of the plurality of risk features according to the feature weights and the corresponding sub-report word counts specifically includes: Calculate the unit word weight corresponding to the risk feature based on the feature weight corresponding to the risk feature and the word count of the sub-report; The multiple risk features are sorted in the first order according to the weight of the number of words per unit. 5. The method as described in claim 1, wherein filtering out at least some risk characteristics based on the first sorting result, the number of words in the sub-report, and predetermined conditions specifically includes: Based on the first sorting result, for each risk feature included in the first sorting result, traverse the list in descending order of weight per unit word count, and execute the following for the current risk feature: The current risk feature is added to the set. It is then determined whether the sum of the word counts of the sub-reports corresponding to the risk features in the set meets the predetermined conditions. If so, the process proceeds to the next risk feature. Otherwise, the current risk feature is removed from the set, the process ends, and the risk features in the set are used as at least a portion of the selected risk features. The set is initially an empty set. 6. The method of claim 5, wherein traversing to the next risk feature specifically includes: Determine the classification accuracy metric of the classification model corresponding to the set of parameters; Determine whether the classification accuracy metric is not greater than the classification accuracy metric of the set corresponding to the classification model before the current risk feature was added; if so, remove the current risk feature from the set and proceed to the next risk feature; otherwise, proceed to the next risk feature. 7. The method of claim 2 or 6, wherein the classification accuracy metric includes the area under the receiver operating characteristic curve (AUC). 8. The method according to any one of claims 1 to 6, wherein the method further comprises: Get the event to be described; For each selected risk feature, a sub-message corresponding to the event to be described is generated. Based on each sub-message, a description message is generated for the event to be described. 9. The method of claim 8, wherein the event to be described is determined by the classification model to be a risk event, and the risk event is a suspected money laundering transaction. 10. A method for describing message generation, comprising: Get the event to be described; Determine the risk characteristics selected from the screening; Based on the selected risk characteristics, a description message is generated for the event to be described; The step of selecting each risk feature includes: obtaining the feature weights of multiple risk features respectively, and selecting each risk feature according to the feature weights and predetermined conditions. The feature weights are obtained based on a classification model trained using sample events or are predefined. The classification model is used to determine risk events, and the predetermined conditions are used to constrain the length of the message generated based on the risk features. Each of the multiple risk features has a corresponding sub-report word count; the process of filtering out each risk feature based on the feature weight and predetermined conditions specifically includes: The multiple risk features are sorted in a first order based on the feature weights and the corresponding number of words in the sub-report; Based on the first sorting result, the number of words in the sub-report, and predetermined conditions, at least some risk characteristics are selected. 11. A risk characteristic screening device, comprising: The acquisition module acquires the feature weights of multiple risk features. The feature weights are obtained based on a classification model trained using sample events or are predefined. The classification model is used to determine risk events. The filtering module filters out at least some risk features based on the feature weights and predetermined conditions, wherein the predetermined conditions are used to constrain the length of the message generated based on the risk features; The plurality of risk features each have a corresponding number of sub-report characters; the filtering module filters out at least a portion of the risk features based on the feature weights and predetermined conditions, specifically including: The filtering module performs a first sorting of the multiple risk features based on the feature weights and the corresponding number of words in the sub-report; Based on the first sorting result, the number of words in the sub-report, and predetermined conditions, at least some risk characteristics are selected. 12. The apparatus of claim 11, further comprising a weight determination module; The weight determination module obtains the feature weights based on the classification model trained using sample events, specifically including: The weight determination module uses sample events to train a classification model; Perform the following actions for each of the aforementioned risk characteristics: Obtain the data corresponding to this risk characteristic from the sample events; Based on the data corresponding to the risk feature, calculate the classification accuracy metric of the classification model corresponding to the risk feature; Based on the classification accuracy metric, the feature weight of this risk characteristic is obtained. 13. The apparatus of claim 11, wherein the screening module performs a first sorting of the plurality of risk features according to the feature weights and the corresponding sub-report word counts, specifically including: The screening module determines the second sorting result obtained by sorting the multiple risk features according to their weights. Based on the second sorting result, at least some of the risk features among the plurality of risk features are selected; The selected risk features are sorted in a first order based on the feature weights and the corresponding number of words in the sub-report. 14. The apparatus of claim 11, wherein the screening module performs a first sorting of the plurality of risk features according to the feature weights and the corresponding sub-report word counts, specifically including: The filtering module calculates the unit word weight corresponding to the risk feature based on the feature weight corresponding to the risk feature and the word count of the sub-report; The multiple risk features are sorted in the first order according to the weight of the number of words per unit. 15. The apparatus of claim 11, wherein the filtering module filters out at least some risk characteristics based on the first sorting result, the number of characters in the sub-report, and predetermined conditions, specifically including: The filtering module, based on the first sorting result, iterates through each risk feature included in the first sorting result in descending order of weight per unit character, and executes the following for the current risk feature: The current risk feature is added to the set. It is then determined whether the sum of the word counts of the sub-reports corresponding to the risk features in the set meets the predetermined conditions. If so, the process proceeds to the next risk feature. Otherwise, the current risk feature is removed from the set, the process ends, and the risk features in the set are used as at least a portion of the selected risk features. The set is initially an empty set. 16. The apparatus of claim 15, wherein the screening module iterates to the next risk feature, specifically including: The filtering module determines the classification accuracy metric of the classification model corresponding to the set; Determine whether the classification accuracy metric is not greater than the classification accuracy metric of the set corresponding to the classification model before the current risk feature was added; if so, remove the current risk feature from the set and proceed to the next risk feature; otherwise, proceed to the next risk feature. 17. The apparatus of claim 12 or 16, wherein the classification accuracy metric includes the area under the receiver operating characteristic curve (AUC). 18. The apparatus according to any one of claims 11 to 16, wherein the apparatus further comprises: The message generation module obtains the event to be described; For each selected risk feature, a sub-message corresponding to the event to be described is generated. Based on each of the sub-messages, a description message is generated for the event to be described. 19. The apparatus of claim 18, wherein the event to be described is determined by the classification model to be a risk event, and the risk event is a suspected money laundering transaction. 20. A message generation apparatus, comprising: Get the module to retrieve the event to be described; The module identifies the selected risk characteristics. The generation module generates a description message for the event to be described based on the selected risk characteristics. The step of selecting each risk feature includes: obtaining the feature weights of multiple risk features respectively, and selecting each risk feature according to the feature weights and predetermined conditions. The feature weights are obtained based on a classification model trained using sample events or are predefined. The classification model is used to determine risk events, and the predetermined conditions are used to constrain the length of the message generated based on the risk features. Each of the multiple risk features has a corresponding sub-report word count; the process of filtering out each risk feature based on the feature weight and predetermined conditions specifically includes: The multiple risk features are sorted in a first order based on the feature weights and the corresponding number of words in the sub-report; Based on the first sorting result, the number of words in the sub-report, and predetermined conditions, at least some risk characteristics are selected. 21. An electronic device for risk characteristic screening, comprising: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, which, when executed by the at least one processor, enable the at least one processor to: Obtain feature weights for multiple risk features. The feature weights are obtained based on a classification model trained using sample events or are predefined. The classification model is used to determine risk events. Based on the feature weights and predetermined conditions, at least some risk features are selected, wherein the predetermined conditions are used to constrain the length of the message generated based on the risk features; The plurality of risk features each have a corresponding number of sub-report characters; the step of selecting at least a subset of risk features based on the feature weights and predetermined conditions specifically includes: The multiple risk features are sorted in a first order based on the feature weights and the corresponding number of words in the sub-report; Based on the first sorting result, the number of words in the sub-report, and predetermined conditions, at least some risk characteristics are selected. 22. An electronic device for describing message generation, comprising: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to: Get the event to be described; Determine the risk characteristics selected from the screening; Based on the selected risk characteristics, a description message is generated for the event to be described; The step of selecting each risk feature includes: obtaining the feature weights of multiple risk features respectively, and selecting each risk feature according to the feature weights and predetermined conditions. The feature weights are obtained based on a classification model trained using sample events or are predefined. The classification model is used to determine risk events, and the predetermined conditions are used to constrain the length of the message generated based on the risk features. Each of the multiple risk features has a corresponding sub-report word count; the process of filtering out each risk feature based on the feature weight and predetermined conditions specifically includes: The multiple risk features are sorted in a first order based on the feature weights and the corresponding number of words in the sub-report; Based on the first sorting result, the number of words in the sub-report, and predetermined conditions, at least some risk characteristics are selected.