CN116364240B - Remote nutrition information processing method and system based on Internet - Google Patents

Abstract

The invention relates to an Internet-based remote nutritional information processing method and system. The system includes a user client, a doctor client and a processing server connected through the Internet. The processing server includes a user database, a plan database, a basic recommendation model and a rapid Recommended model. The processing server selects the corresponding recommendation model and generates recommended nutrition plans based on the personal information uploaded by the user client, which can provide fast service response for large-scale users on the Internet.

Description

An Internet-based remote nutritional information processing method and system

【Technical field】

The invention belongs to the field of telemedicine, and in particular relates to an Internet-based remote nutritional information processing method and system.

【Background technique】

The hospital nutritionist will conduct a nutritional assessment on the patient based on the patient's personal situation, such as height, weight, dietary intake, current illness, past medical history, etc., and provide personalized nutrition guidance based on the assessment, and provide appropriate nutritional support program.

With the rise of the Internet, patients can now be diagnosed and treated remotely through the Internet. Relying on Internet telemedicine, nutritionists can provide personalized nutrition guidance not only to patients, but also to ordinary users. However, due to the large number of Internet users, there is a huge demand for nutritional guidance, and the number of doctors participating in remote nutritional guidance is small, making it difficult to meet the large-scale demand for the Internet. To this end, there are currently some methods of analyzing and recommending nutritional guidance programs through computers. That is, the computer runs a recommendation model to analyze the personal information provided by the user, recommends a nutritional guidance program based on the personal information, and then the doctor analyzes the recommended nutrition guidance program. The plan can be reviewed and adjusted, thereby reducing the workload of doctors and allowing them to serve more users.

However, if the existing recommendation model needs to make accurate recommendations, its calculation time is usually relatively long, requiring users to wait for a long time, and the user experience is poor.

[Content of the invention]

In order to solve the above-mentioned problems in the prior art, the present invention provides an Internet-based remote nutrition information processing method and system, which can quickly recommend nutrition guidance programs based on the user's personal information.

The technical solutions adopted by the present invention are as follows:

An Internet-based remote nutrition information processing method includes the following steps:

Step 1: The user uses the user client to log in to the processing server and upload personal information;

Step 2: The processing server converts the personal information into the corresponding personal information vector V _now ;

Step 3: The processing server determines whether a nutritional guidance program has been recommended to the user before;

Step 4: If the processing server has not recommended a nutritional guidance program to the user before, the processing server inputs the personal information vector V _now into the basic recommendation model, and the basic recommendation model outputs the corresponding response based on the input personal information vector. recommended solution; then jump to step 7;

Step 5: If the processing server has recommended a nutritional guidance program to the user, the processing server queries the user's last recommendation record from the user database, and obtains the corresponding personal information vectors V _last and V from the recommendation record. Recommended solution identifier ProjectID _last ;

Step 6: The processing server calculates the similarity between the current personal information vector V _now and the last personal information vector V _last ; if the similarity is less than a predetermined threshold, the processing server inputs the personal information vector V _now The basic recommendation model is used to obtain the corresponding recommendation solution. If the similarity is greater than or equal to the predetermined threshold, the processing server inputs the personal information vector V _now and the last recommendation solution identifier ProjectID _last into the fast recommendation model to obtain the corresponding recommendation solution. recommended solutions;

Step 7: The processing server sends the personal information and the corresponding recommended plan to the doctor client, and the doctor reviews and adjusts the recommended plan on the doctor client to generate the final plan.

Further, the user database is used to store the user's registration information and the user's recommendation record; the recommendation record includes the recommendation time, the user's personal information vector, and the recommendation solution identifier obtained by the processing server based on the personal information vector. .

Further, the program database stores multiple pre-designed nutrition guidance programs, each program has at least one corresponding personal information vector in advance, and the identifier of each program in the program database and its corresponding at least one personal information are The vector is used as a training sample to train the basic recommendation model; the personal information vector is input into the trained basic recommendation model to output the identifier of the corresponding recommendation solution.

Further, the fast recommendation model is a pre-trained recommendation model, which inputs the user's current personal information vector and the identifier of the user's last recommendation solution, and outputs the identifier of the recommendation solution currently suitable for the user.

Furthermore, both the basic recommendation model and the fast recommendation model use deep neural network models, and the number of neural network layers of the fast recommendation model is less than the number of layers of the basic recommendation model.

Further, obtaining the training samples of the fast recommendation model through the calculation results of the basic recommendation model includes: obtaining two personal information vectors V ₁ and V ₂ whose similarity is greater than or equal to a predetermined threshold, and then obtaining the corresponding vector of V ₁ through the basic recommendation model. Recommendation scheme identifier ID ₁ , and recommendation scheme identifier ID ₂ corresponding to V ₂ , use (V ₁ , ID ₂ , tag ID ₁ ) and (V ₂ , ID ₁ , tag ID ₂ ) as training samples for the fast recommendation model .

Further, the processing server records the final solution and the corresponding personal information vector into the solution database for retraining the basic recommendation model.

Further, the doctor client returns the final solution to the processing server, and the processing server returns the final solution to the user client.

The invention also provides an Internet-based remote nutrition information processing system, which is used to implement the above method. The system includes a user client, a doctor client and a processing server connected through the Internet. The processing server includes a user database. , solution database, basic recommendation model and fast recommendation model, wherein the user database is used to store the user's registration information and the user's recommendation record; the recommendation record includes the recommendation time, the user's personal information vector, and the processing server based on The recommended program identifier obtained from the personal information vector; the program database stores multiple pre-designed nutrition guidance programs, each program has at least one corresponding personal information vector in advance, and each program has a corresponding unique identifier; so The basic recommendation model inputs a personal information vector and outputs the identifier of the corresponding recommendation plan; the fast recommendation model is a pre-trained recommendation model that inputs the user's current personal information vector and the identifier of the user's last recommendation plan. , output the recommended solution identifier currently suitable for the user.

Further, the user client is a personal computer, a smart phone or a tablet computer, the doctor client is a personal computer, a smart phone or a tablet computer, and the processing server is a single server or a server cluster.

The beneficial effects of the present invention are: providing nutritional guidance programs for ordinary users on the Internet. When the number of nutritionists is small, the present invention can also provide faster service response for large-scale users on the Internet.

[Picture description]

The accompanying drawings described here are used to provide a further understanding of the present invention and constitute a part of this application, but do not constitute an improper limitation of the present invention. In the accompanying drawings:

Figure 1 is a basic structural diagram of the remote nutrition information processing system of the present invention.

【Detailed ways】

The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. The illustrative embodiments and descriptions are only used to explain the present invention, but are not intended to limit the present invention.

Referring to Figure 1, it shows the basic structure of the remote nutritional information processing system of the present application. The system includes a user client, a processing server and a doctor client. Various clients and servers are connected and communicate with each other through the Internet.

Among them, the user client is a client used by Internet users. Users can use their user clients to log in to the processing server in the system, communicate with the processing server, including filling in and uploading personal information, and receiving nutritional guidance plans returned by the processing server. . Although only one user client is shown in Figure 1, those skilled in the art can understand that the system may include multiple different user clients used by multiple users. The specific user client can be a personal computer, a smartphone, a tablet, etc.

The processing server is the core processing device of this application. It can receive personal information uploaded by logged-in users, obtain the recommended nutritional guidance program through analysis of the personal information, and notify the doctor of the recommended nutritional guidance program (hereinafter referred to as the recommended program). The client receives the final nutritional guidance plan (hereinafter referred to as the final plan) returned by the doctor client, and returns the final plan to the user client. The processing server also includes a user database, a plan database, a basic recommendation model, and a fast recommendation model. The specific functions of the user database, plan database, basic recommendation model, and fast recommendation model will be explained later. It should be noted here that the model may be a recommendation model implemented by software programming, a recommendation model implemented by hardware, or a recommendation model implemented by a combination of software and hardware. The present invention does not limit this. In specific implementation, the processing server may be a single server or a server cluster, which is not limited by the present invention.

The doctor client is a client used by nutritionists. Doctors can use the doctor client to log in to the above-mentioned processing server, receive personal information and recommended plans submitted by the processing server, and return the final plan reviewed and adjusted by the doctor to the processing server. Although only one doctor client is shown in Figure 1, those skilled in the art can understand that the system may include multiple different doctor clients used by multiple doctors. The specific doctor client can be a personal computer, a smartphone, a tablet, etc.

Based on the basic structure of the above system, the remote nutrition information processing method of the present application will be described in detail below.

Step 1: The user uses the user client, logs in to the processing server, and uploads personal information.

Specifically, the user needs to register in the system first, and then use the registration information to log in to the processing server. For example, the user can use the registered account and password to log in to the processing server. Registration and login are existing technologies in this field and will not be described again here.

After the user logs in to the processing server, if he or she needs to obtain his or her own nutritional guidance plan, he or she must first fill in his or her personal information using the user client. The specific items of the personal information can be determined in advance by the system, such as gender, date of birth, height, weight, blood pressure, dietary preferences, bad habits, past medical history, family medical history, etc. Preferably, the user client can collect personal information in the form of a questionnaire. The questionnaire is formulated in advance by a nutritionist or system administrator. The user fills out the questionnaire in the user client and answers each question in the questionnaire (can It is a multiple-choice question or a fill-in-the-blank question), so that the user client can obtain the user's personal information based on the answers filled in by the user. After the user completes filling in the personal information, the user client submits the personal information to the processing server.

Step 2: The processing server converts the personal information into the corresponding personal information vector V _now .

Before recommending the model for data analysis, personal information needs to be processed accordingly and converted into corresponding vector data. For data with fixed options, such as gender data, you can directly convert it to a specified value as a vector value, such as converting "male" to 1 and "female" to 0. For data with corresponding values, you can use the values directly. For example, calculate the user's current age based on the year and month of birth as a vector value. For personal data in the form of text, it is necessary to clean the data and remove invalid data, and then convert the text or keywords into vector values through a predetermined method. There are many existing methods in this field that can convert text into vectors, such as Word2vec etc., the present invention does not limit which method is specifically used. If a piece of data is missing from the uploaded personal information, its corresponding vector value can be set to 0.

In short, after processing the personal information, the processing server can convert it into an n-dimensional personal information vector V _now =<P ₁ , P ₂ ,..., P _n >.

Step 3: The processing server determines whether a nutritional guidance program has been recommended to the user before.

Specifically, if the user requests nutritional guidance using the system of the present invention for the first time, the processing server may determine that no nutritional guidance program has been recommended for the user before, and that this is the first time that the nutritional guidance program is recommended for the user. If the user has used the system of the present invention to obtain a nutritional guidance program before, there will be a corresponding recommendation record in the user database of the processing server, so that the processing server can determine that a nutritional guidance program has been recommended for the user.

The user database of the processing server is used to store the user's registration information and the user's recommendation records. That is, every time a user requests and obtains a nutritional guidance plan through the processing server, the user database will store a recommendation record, and the recommendation record includes the personal information vector of the personal information uploaded by the user at that time, and the nutritional guidance recommended by the processing server at that time. The identifier of the scheme (e.g. scheme number). By querying the user database, the processing server can determine whether the user requests nutritional guidance for the first time and whether a nutritional guidance program has been recommended to the user before.

Step 4: If the processing server has not recommended a nutritional guidance program to the user before, the processing server inputs the personal information vector V _now into the basic recommendation model, and the basic recommendation model outputs the corresponding response based on the input personal information vector. recommended solution; then jump to step 7.

Specifically, when the processing server determines that the user has not recommended a nutrition guidance program before, that is, when the user requests nutrition guidance for the first time, it calls the basic recommendation model to complete the recommendation of the nutrition guidance program. The basic recommendation model is a pre-trained recommendation model, which inputs the user's personal information vector and outputs a recommendation solution suitable for the user. The specific recommendation model can use any existing recommendation model in the field, and the present invention does not limit this. Preferably, the basic recommendation model can adopt a deep neural network model.

In a specific implementation, the nutritionist can consider different personal situations (i.e., different personal information vectors) and design corresponding nutritional guidance programs for each personal situation in advance, thereby obtaining multiple nutritional guidance programs and combining all previously designed programs. Stored in the solution database of the processing server, that is to say, each solution in the solution database has at least one corresponding personal information vector in advance. Each scenario can have its corresponding unique identifier, such as a unique number for the scenario. Therefore, the identifier of each plan in the plan database and its corresponding at least one personal information vector are used as training samples, and the basic recommendation model is trained to obtain a trained basic recommendation model. In this way, the personal information vector is input into the trained basic recommendation model, and the model can output the identifier of the recommended program, and the processing server queries the program database to obtain the specific nutritional guidance program based on the identifier.

Step 5: If the processing server has recommended a nutritional guidance program to the user, the processing server queries the user's last recommendation record from the user database, and obtains the corresponding personal information vectors V _last and V from the recommendation record. Recommended project identifier ProjectID _last .

As mentioned above, the user database stores each recommendation record of the user, including the recommendation time of the recommendation record, the personal information vector calculated based on the personal information uploaded by the user at that time, and the recommendation obtained by the processing server based on the personal information vector. Protocol identifier; may also include the final protocol resulting from the physician's review and adjustment of the recommended protocol. Therefore, the processing server can query and obtain the user's last recommendation record from the user database, and can obtain the user's last personal information vector and the last recommendation plan identifier from the recommendation record.

Step 6: The processing server calculates the similarity between the current personal information vector V _now and the last personal information vector V _last ; if the similarity is less than a predetermined threshold, the processing server inputs the personal information vector V _now The basic recommendation model is used to obtain the corresponding recommendation solution. If the similarity is greater than or equal to the predetermined threshold, the processing server inputs the personal information vector V _now and the last recommendation solution identifier ProjectID _last into the fast recommendation model to obtain the corresponding recommendation solution. recommended solution.

The processing server selects whether to use the basic recommendation model or the fast recommendation model based on the similarity between V _now and V _last . That is to say, if the personal information changes significantly, the original recommendation plan is of little reference, and the basic recommendation model needs to be reused for recommendation. If the personal information does not change much compared to the last time, the original recommendation plan The reference is greater, and the recommended solution may even remain unchanged. In this case, the fast recommendation model can be used for recommendation. In actual use, in many cases personal information does not change much compared to the last time. It may only be age, weight and other items that have changed. Therefore, a fast recommendation model can be applied. The fast recommendation model makes a quick recommendation based on the last recommendation plan. Recommended to improve the overall operating efficiency of the system.

The similarity can be calculated using any existing similarity calculation method. For example, cosine similarity can be used as the similarity between two vectors. The present invention does not limit the specific similarity algorithm.

The fast recommendation model is a pre-trained recommendation model, which inputs the user's current personal information vector and the identifier of the last recommendation plan, and outputs the current recommendation plan identifier suitable for the user.

Similar to the basic recommendation model, the fast recommendation model can also adopt any existing recommendation model in the field, and the present invention is not limited to this. However, the computational efficiency of the fast recommendation model should be greatly improved compared to the basic recommendation model. For example, in a preferred embodiment of the present invention, both the basic recommendation model and the fast recommendation model adopt a deep neural network model, but the number of layers of the fast recommendation model is less than the number of layers of the basic recommendation model. For example, the basic recommendation model can There are n hidden layers, while the fast recommendation model only has n/2 hidden layers. In this way, the calculation scale of the fast recommendation model is smaller than the basic recommendation model and the calculation efficiency is high.

The training samples of the fast recommendation model can be obtained from the calculation results of the basic recommendation model. For example, obtain two personal information vectors V ₁ and V ₂ whose similarity is greater than or equal to a predetermined threshold, and then obtain the recommendation solution identifier ID ₁ corresponding to V ₁ and the recommendation solution identifier ID ₂ corresponding to V ₂ through the basic recommendation model. Then (V ₁ , ID ₂ , tag ID ₁ ) and (V ₂ , ID ₁ , tag ID ₂ ) can both be used as training samples for the fast recommendation model.

Although the fast recommendation model is relatively simpler than the basic recommendation model, because the fast recommendation model uses the calculation results of the basic recommendation model, the accuracy, precision and recall of the trained fast recommendation model can only be slightly higher than that of the basic recommendation model. There has been a decrease, but the calculation speed has been greatly improved, which is generally conducive to the promotion and use of the system.

Step 7: The processing server sends the personal information and the corresponding recommended plan to the doctor client, and the doctor reviews and adjusts the recommended plan on the doctor client to generate the final plan.

For existing machine recommendation models, the output results are not guaranteed to be completely correct. The generated recommendation plan has a small probability of errors, so the final plan needs to be reviewed by real doctors. Specifically, the processing server determines the corresponding recommendation solution, that is, the recommendation solution identifier, in the above-mentioned step 4 or step 6, and by querying the solution database, the processing server can obtain the specific recommendation solution content. The processing server then sends the specific recommendation plan content, together with the user's personal information, to the doctor client for review and adjustment by the doctor.

The doctor uses the doctor client to review the user's current personal information and the specific recommended plan content. If necessary, the specific nutritional guidance content can be adjusted to form the final plan. The doctor client returns the final solution to the processing server, and the processing server returns the final solution to the user client. The processing server can also save the final solution in the user database for subsequent query. If the specific contents of the final solution and the recommended solution are different, the processing server can also record the final solution and the corresponding personal information vector into the solution database for subsequent training of the recommendation model.

Through the above method steps and system of the present invention, the present invention can provide nutrition guidance programs for ordinary users on the Internet. When the number of nutritionists is small, the present invention can also provide faster nutrition guidance programs for large-scale users on the Internet. Service response.

The above are only preferred embodiments of the present invention. Therefore, any equivalent changes or modifications based on the structures, features and principles described in the patent application scope of the present invention are included in the patent application scope of the present invention.

Claims (9)

1. An Internet-based remote nutrition information processing method, characterized by including the following steps: Step 1: The user uses the user client to log in to the processing server and upload personal information; Step 2: The processing server converts the personal information into the corresponding personal information vector V _now ; Step 3: The processing server determines whether a nutritional guidance program has been recommended to the user before; Step 4: If the processing server has not recommended a nutritional guidance program to the user before, the processing server inputs the personal information vector V _now into the basic recommendation model, and the basic recommendation model outputs the corresponding response based on the input personal information vector. recommended solution; then jump to step 7; Step 5: If the processing server has recommended a nutritional guidance program to the user, the processing server queries the user's last recommendation record from the user database, and obtains the corresponding personal information vectors V _last and V from the recommendation record. Recommended solution identifier ProjectID _last ; Step 6: The processing server calculates the similarity between the current personal information vector V _now and the last personal information vector V _last ; if the similarity is less than a predetermined threshold, the processing server inputs the personal information vector V _now The basic recommendation model is used to obtain the corresponding recommendation solution. If the similarity is greater than or equal to the predetermined threshold, the processing server inputs the personal information vector V _now and the last recommendation solution identifier ProjectID _last into the fast recommendation model to obtain the corresponding recommendation solution. recommended solutions; Step 7: The processing server sends the personal information and the corresponding recommended plan to the doctor client, and the doctor reviews and adjusts the recommended plan on the doctor client to generate the final plan. 2. The method according to claim 1, characterized in that the user database is used to store the user's registration information and the user's recommendation record; the recommendation record includes recommendation time, the user's personal information vector, and processing The recommended solution identifier obtained by the server based on the personal information vector. 3. The method according to any one of claims 1-2, characterized in that the program database stores a plurality of nutritional guidance programs designed in advance, each program has at least one corresponding personal information vector in advance, and the program database stores The identifier of each plan and its corresponding at least one personal information vector are used as training samples to train the basic recommendation model; the personal information vector is input into the trained basic recommendation model to output the identifier of the corresponding recommendation plan. 4. The method according to any one of claims 1-2, characterized in that the fast recommendation model is a pre-trained recommendation model, which inputs the user's current personal information vector and the user's last recommendation plan. identifier, output the identifier of the recommended solution currently suitable for the user. 5. The method according to any one of claims 1-2, characterized in that both the basic recommendation model and the fast recommendation model adopt a deep neural network model, and the number of neural network layers of the fast recommendation model is less than the number of layers of the basic recommendation model. . 6. The method according to claim 1, characterized in that the processing server records the final solution and the corresponding personal information vector into the solution database for retraining the basic recommendation model. 7. The method according to claim 1, characterized in that the doctor client returns the final solution to the processing server, and the processing server returns the final solution to the user client. 8. A remote nutritional information processing system based on the Internet, characterized in that the system is used to implement the method of claim 1, and the system includes a user client, a doctor client and a processing server connected through the Internet, so The above processing server includes user database, solution database, basic recommendation model and fast recommendation model, among which The user database is used to store the user's registration information and the user's recommendation record; the recommendation record includes the recommendation time, the user's personal information vector, and the recommendation plan identifier obtained by the processing server based on the personal information vector; The program database stores multiple pre-designed nutrition guidance programs, each program has at least one corresponding personal information vector in advance, and each program has a corresponding unique identifier; The basic recommendation model inputs a personal information vector and outputs the identifier of the corresponding recommendation plan; the fast recommendation model is a pre-trained recommendation model that inputs the user's current personal information vector and the identifier of the user's last recommendation plan. symbol, output the identifier of the recommended solution currently suitable for the user. 9. The system according to claim 8, wherein the user client is a personal computer, a smart phone or a tablet, the doctor client is a personal computer, a smart phone or a tablet, and the processing server is A single server or a cluster of servers.