Health management platform based on deep learning

Abstract

The invention discloses a health management platform based on deep learning, and belongs to the technical field of data processing, the platform can perform qualification verification and real-name authentication on registered doctors through setting a doctor registration module, so as to ensure the specialty of the doctors and provide basic guarantee for subsequent automatic health management; meanwhile, a physician data interaction module is arranged to collect historical user body data and corresponding health management suggestions, so that data for deep learning is obtained; then, a deep learning module is adopted to learn historical user body data and corresponding health management suggestions, and health guidance rules are formed; and finally, a health guidance rule is scheduled through the health guidance module to analyze the current user body data input by the user data interaction module, so that a target health management suggestion is obtained, the health management efficiency can be effectively improved, and the user can better manage own health and prevent diseases.

Description

Technical Field

[0001] The present invention belongs to the field of data processing technology, and specifically relates to a health management platform based on deep learning. Background Art

[0002] At present, with the accelerated pace of life, people pay more and more attention to their own health management. However, traditional health management methods often rely on manual statistics and analysis, which are inefficient and inaccurate. Although there are some Internet-based health management platforms, they still have a lot of room for improvement in data processing and analysis. Therefore, it is particularly important to develop a platform that can efficiently and accurately analyze health data.

[0003] In existing health management platforms, users often input data and then receive health management guidance from professional doctors. On the one hand, this requires a high level of professional knowledge, and on the other hand, the health management efficiency is low, and it is often impossible to provide timely advice to users. Summary of the invention

[0004] The present invention provides a health management platform based on deep learning, which is used to solve the technical problems in the prior art that the health management efficiency is low and it is often impossible to provide timely suggestions to users.

[0005] A health management platform based on deep learning, comprising: a physician registration module, a physician data interaction module, a deep learning module, a user data interaction module and a health guidance module; The physician registration module is used to obtain the registration request transmitted by the physician, and after reviewing the registration request, generate the physician account and password; The physician data interaction module is used to verify the physician account and password entered by the physician, and obtain the historical user physical data and corresponding health management suggestions entered by the physician after the verification is passed; The deep learning module is used to use a deep learning algorithm to perform data learning on the historical user physical data input by the physician and the corresponding health management suggestions to obtain health guidance rules; The user data interaction module is used to obtain the current user body data input by the user through human-computer interaction, and provide the current user body data to the health guidance module; The health guidance module is used to call the health guidance rules to process the current user's physical data, obtain target health management suggestions, and present the target health management suggestions to the user through a visual interface.

[0006] Furthermore, the registration request transmitted by the physician is obtained, and after reviewing the registration request, a physician account and password are generated, including: Obtaining a registration request transmitted by a physician; wherein the registration request includes an ID number, a mobile phone number, a portrait image, an ID image, and a physician's certificate; Performing joint identity authentication using a portrait image, an ID card image, and a physician's certificate to obtain a first authentication result; wherein the first authentication result includes authentication success or authentication failure; Conducting an online query on the qualification certificate number on the physician certificate to determine a second verification result corresponding to the physician certificate; wherein the second verification result includes verification success or verification failure; Performing unified verification of the ID card number and the ID card image to determine a third verification result corresponding to the ID card; wherein the third verification result includes verification success or verification failure; When the first verification result, the second verification result and the third verification result are all successful, a unique physician account and password are randomly generated.

[0007] Furthermore, historical user physical data includes: height, weight, gender, age, body mass index, waist circumference, waist-to-hip ratio, body fat percentage, and physical function assessment data.

[0008] Furthermore, after obtaining the historical user physical data and corresponding health management suggestions input by the physician, the method further includes: Determine the currently input health management suggestion, extract entities or keywords of the health management suggestion, and perform text similarity matching of the entities or keywords in the existing health management suggestions to obtain the target health management suggestion; Determine whether the text similarity between the currently input health management suggestion and the target health management suggestion is greater than a preset threshold, if so, determine that the currently input health management suggestion is a possible duplicate suggestion, otherwise determine that the currently input health management suggestion is a non-duplicate suggestion; When the currently input health management suggestion is a possible duplicate suggestion, the currently input health management suggestion is randomly transmitted to another physician for secondary judgment to obtain a secondary judgment result; wherein the secondary judgment result includes whether the currently input health management suggestion is a duplicate suggestion or a non-duplicate suggestion; When the secondary judgment result includes that the currently input health management suggestion is a duplicate suggestion, the health management suggestion corresponding to the historical user physical data input by the currently input physician is replaced with the target health management suggestion; When the secondary judgment result includes that the currently input health management suggestion is not a duplicate suggestion, the health management suggestion corresponding to the historical user physical data currently input by the physician is stored in the database; When the currently input health management suggestion is not a duplicate suggestion, the currently input health management is stored in the database.

[0009] Furthermore, a deep learning algorithm is used to learn the historical user physical data and corresponding health management suggestions input by the physician to obtain health guidance rules, including: Constructing a corresponding unique classification number for each health management suggestion, and associating the unique classification number with the historical user physical data input by the physician; A deep learning algorithm is used to construct the initial rules, and the parameters of the initial rules are updated through the historical user physical data with association and the unique classification number corresponding to the historical user physical data, so that the initial rules are adjusted to the data relationship between the historical user physical data and the unique classification number to obtain health guidance rules.

[0010] Furthermore, the parameters of the initial rule are updated by using the historical user physical data having an associated relationship and the unique classification number corresponding to the historical user physical data, including: The parameters of the initial rule are randomly initialized, and the parameters of the initial rule are encoded using a vector encoding method to obtain an encoded individual; Repeat the process of obtaining the coded individuals multiple times to obtain multiple different coded individuals; After applying the parameters contained in the coded individual to the initial rule, the historical user body data is used as the input of the initial rule, and the unique classification number corresponding to the historical user body data is used as the expected output to obtain the fitness value corresponding to each coded individual; Based on the fitness value corresponding to each coding individual, an adaptive update step is determined, and a nearby area of ​​the coding individual is searched according to the adaptive update step to determine the coding individual after a small range search; Based on the fitness value corresponding to each coding individual, the optimal coding individual is determined, and the position information exchange individual is determined based on the optimal coding individual. The position information exchange individual is used to further search the coding individuals after the small range search to determine the coding individuals after the extended search; Randomly matching the coded individual after the extended search with another coded individual after the extended search to obtain the matching coded individual corresponding to each coded individual after the extended search, and performing an unknown area search between the coded individuals after the extended search according to the matching coded individuals to obtain the coded individuals after the unknown area search; Determine the historical optimal state corresponding to each coding individual after the unknown area search, and perform an irregular search on the coding individual after the unknown area search according to the historical optimal state to determine the coding individual after the irregular search; Determine whether the current update number is greater than the preset maximum update number. If so, re-determine the optimal coding individual based on the coding individuals after the irregular search. Otherwise, return to the step of obtaining the fitness value corresponding to each coding individual.

[0011] Further, based on the fitness value corresponding to each coding individual, an adaptive update step is determined, and a nearby area of ​​the coding individual is searched according to the adaptive update step to determine the coding individual after the small range search, including: Based on the fitness value corresponding to each coded individual, determine the fitness ratio corresponding to the coded individual; Taking the fitness ratio corresponding to the encoded individuals as the power of the power function, and combining the update step in the previous update process, adaptively determining the update step in the current update process, and obtaining an adaptive update step; Based on the adaptive update step size, different amounts of search are performed for the parameters of each dimension in the coding individual to achieve a search in a nearby area and obtain the coding individual after a small-range search.

[0012] Further, based on the fitness value corresponding to each coding individual, the optimal coding individual is determined, and the position information exchange individual is determined according to the optimal coding individual, and the coding individual after the small range search is further searched using the position information exchange individual to determine the coding individual after the extended search, including: Based on the fitness value corresponding to each coding individual, the optimal coding individual is determined; For any coded individual after searching in a small range, the coded individual is randomly matched with a first random coded individual and a second random coded individual; After exchanging information between the first random coding individual and the second random coding individual, an adaptive update step size is used for control, and the position information exchange individual is obtained in combination with the position of the optimal coding individual; After using Gaussian distribution to perturb the position information exchange individuals, the first search item corresponding to the coded individuals after a small range search is obtained; The first search item is used to perform an extended search on the coded individuals after the narrow range search to obtain coded individuals after the extended search.

[0013] Furthermore, the coded individuals after the extended search are randomly matched with another coded individual after the extended search to obtain matching coded individuals corresponding to each coded individual after the extended search, and unknown area search is performed between individuals on the coded individuals after the extended search according to the matching coded individuals to obtain coded individuals after the unknown area search, including: Randomly match the coded individual after the extended search with another coded individual after the extended search, and obtain the matching coded individual corresponding to each coded individual after the extended search; Performing information exchange between the coded individual after the extended search and the corresponding matching coded individual to obtain an information exchange individual; Based on the current update number, the information exchange individuals are adaptively searched and controlled using normal distribution to obtain a second search item corresponding to the coding individual after the extended search; The information exchange individual is added to the second search item corresponding to the coding individual to determine the coding individual after the unknown area search.

[0014] Further, determining the historical optimal state corresponding to each coding individual after the unknown area search, and performing an irregular search on the coding individual after the unknown area search according to the historical optimal state, and determining the coding individual after the irregular search, including: Determine the historical optimal state corresponding to the coding individual after searching each unknown area; For any coded individual after searching in an unknown area, a Lvey flight step is generated for it through the Lvey flight strategy; The Lvey flight step length is rounded to obtain the irregular information fusion quantity, and according to the irregular information fusion quantity, a plurality of irregular decision individuals are paired for the coding individual; According to the historical optimal states corresponding to the plurality of irregular decision individuals and the coding individuals, the coding individuals are irregularly searched to determine the irregular search individuals; Determine whether the fitness value of the irregular search individual is greater than the fitness value corresponding to the coded individual after the unknown area search. If so, use the irregular search individual to update the coded individual after the unknown area search to obtain the coded individual after the irregular search. Otherwise, directly use the original coded individual after the unknown area search as the coded individual after the irregular search.

[0015] The present invention provides a health management platform based on deep learning. The platform can verify the qualifications and real-name authentication of registered doctors by setting up a physician registration module to ensure the professionalism of the doctors and provide basic protection for subsequent automatic health management; at the same time, a physician data interaction module is set up to collect historical user body data and corresponding health management suggestions, so as to obtain data for deep learning; then the deep learning module is used to learn the historical user body data and the corresponding health management suggestions to form health guidance rules; finally, the health guidance module schedules the health guidance rules to analyze the current user body data input by the user data interaction module, so as to obtain target health management suggestions, which can effectively improve the efficiency of health management and help users better manage their own health and prevent diseases. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0017] Figure 1 A schematic diagram of the structure of a health management platform based on deep learning provided in an embodiment of the present invention.

[0018] Figure 2 A schematic diagram of a process for updating parameters of an initial rule provided by an embodiment of the present invention.

[0019] Among them, 101-physician registration module, 102-physician data interaction module, 103-deep learning module, 104-user data interaction module, and 105-health guidance module.

[0020] The above drawings have shown clear embodiments of the present invention, which will be described in more detail below. These drawings and text descriptions are not intended to limit the scope of the present invention in any way, but to illustrate the concept of the present invention for those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0021] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Instead, they are merely examples of devices and methods consistent with some aspects of the present invention as detailed in the appended claims.

[0022] The embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0023] like Figure 1 As shown, an embodiment of the present invention provides a health management platform based on deep learning, including: a physician registration module 101, a physician data interaction module 102, a deep learning module 103, a user data interaction module 104 and a health guidance module 105; The physician registration module 101 is used to obtain a registration request transmitted by a physician, and after reviewing the registration request, generate a physician account and password; The registration request may include various identity information and qualification certificates of the physician. By verifying these identity information and qualification certificates, it is determined that the physician is professional and can provide accurate advice, providing a strong foundation for subsequent automatic advice.

[0024] The physician data interaction module 102 is used to verify the physician account and password input by the physician, and obtain the historical user physical data and corresponding health management suggestions input by the physician after the verification is passed; Historical user physical data can be obtained from the user or through some databases, and then displayed to the physician to obtain health management suggestions input by the physician, thereby forming different data pairs to facilitate the subsequent formation of data recognition rules.

[0025] The deep learning module 103 is used to use a deep learning algorithm to perform data learning on the historical user physical data input by the physician and the corresponding health management suggestions to obtain health guidance rules; A deep learning algorithm may refer to a neural network algorithm. After learning historical user physical data and corresponding health management recommendations through the neural network algorithm, a health guidance rule can be obtained, thereby realizing automated data recognition and recommendations.

[0026] The user data interaction module 104 is used to obtain the current user body data input by the user through human-computer interaction, and provide the current user body data to the health guidance module 105; The data types included in the current user's physical data are the same as the data types included in the historical user's physical data to ensure that the data can be accurately identified.

[0027] The health guidance module 105 is used to call the health guidance rules to process the current user's physical data, obtain target health management suggestions, and present the target health management suggestions to the user through a visual interface.

[0028] The present invention provides a health management platform based on deep learning. The platform can verify the qualifications and real-name authentication of registered doctors by setting up a physician registration module to ensure the professionalism of the doctors and provide basic protection for subsequent automatic health management; at the same time, a physician data interaction module is set up to collect historical user body data and corresponding health management suggestions, so as to obtain data for deep learning; then the deep learning module is used to learn the historical user body data and the corresponding health management suggestions to form health guidance rules; finally, the health guidance module schedules the health guidance rules to analyze the current user body data input by the user data interaction module, so as to obtain target health management suggestions, which can effectively improve the efficiency of health management and help users better manage their own health and prevent diseases.

[0029] In the embodiment of the present invention, the registration request transmitted by the physician is obtained, and after the registration request is reviewed, the physician account and password are generated, including: Obtaining a registration request transmitted by a physician; wherein the registration request includes an ID number, a mobile phone number, a portrait image, an ID image, and a physician's certificate; A joint identity verification is performed using a portrait image, an ID card image, and a physician's certificate to obtain a first verification result; wherein the first verification result includes a successful verification or a failed verification; the first verification result can ensure that the physician himself has registered, thereby ensuring the accuracy of the recommendation.

[0030] The qualification certificate number on the physician's certificate is queried online to determine a second verification result corresponding to the physician's certificate; wherein the second verification result includes verification success or verification failure; the second verification result can ensure whether the physician's qualifications are authentic, thereby ensuring the accuracy of the advice.

[0031] Performing unified verification of the ID card number and the ID card image to determine a third verification result corresponding to the ID card; wherein the third verification result includes verification success or verification failure; When the first verification result, the second verification result and the third verification result are all successful, a unique physician account and password are randomly generated.

[0032] It is worth noting that in addition to the above-mentioned verifications, other verifications can be used to further ensure that the advice given by the physician is accurate.

[0033] In the embodiment of the present invention, the historical user physical data includes: height, weight, gender, age, body mass index, waist circumference, waist-to-hip ratio, body fat percentage, and physical function evaluation data.

[0034] Optionally, the physical function assessment data can be heart rate, blood pressure, respiratory rate and other body-related data over a period of time, so that more accurate suggestions can be given. Health management suggestions can be dietary suggestions, work and rest suggestions, or exercise suggestions.

[0035] In the embodiment of the present invention, after obtaining the historical user physical data input by the physician and the corresponding health management suggestions, the method further includes: Determine the currently input health management suggestion, extract entities or keywords of the health management suggestion, and perform text similarity matching of the entities or keywords in the existing health management suggestions to obtain the target health management suggestion; Determine whether the text similarity between the currently input health management suggestion and the target health management suggestion is greater than a preset threshold, if so, determine that the currently input health management suggestion is a possible duplicate suggestion, otherwise determine that the currently input health management suggestion is a non-duplicate suggestion; When the currently input health management suggestion is a possible duplicate suggestion, the currently input health management suggestion is randomly transmitted to another physician for secondary judgment to obtain a secondary judgment result; wherein the secondary judgment result includes whether the currently input health management suggestion is a duplicate suggestion or a non-duplicate suggestion; When the secondary judgment result includes that the currently input health management suggestion is a duplicate suggestion, the health management suggestion corresponding to the historical user physical data input by the currently input physician is replaced with the target health management suggestion; When the secondary judgment result includes that the currently input health management suggestion is not a duplicate suggestion, the health management suggestion corresponding to the historical user physical data currently input by the physician is stored in the database; When the currently input health management suggestion is not a duplicate suggestion, the currently input health management is stored in the database.

[0036] By removing duplicate recommendations from different physicians, the number of final classifications can be effectively reduced, thereby reducing complexity.

[0037] In the embodiment of the present invention, a deep learning algorithm is used to learn the historical user physical data and corresponding health management suggestions input by the physician to obtain health guidance rules, including: Constructing a corresponding unique classification number for each health management suggestion, and associating the unique classification number with the historical user physical data input by the physician; A deep learning algorithm is used to construct the initial rules, and the parameters of the initial rules are updated through the historical user physical data with association and the unique classification number corresponding to the historical user physical data, so that the initial rules are adjusted to the data relationship between the historical user physical data and the unique classification number to obtain health guidance rules.

[0038] The deep learning algorithm can be set as a convolutional neural network (CNN) or a long short-term memory network (LSTM), or as a CNN-LSTM, so as to achieve spatial and temporal analysis of the data. The parameters of the deep learning algorithm are mainly weights and thresholds. Therefore, after the initial rules are constructed using the deep learning algorithm, these hyperparameters can be updated so that the final health guidance rules can make accurate predictions.

[0039] It is worth noting that the health guidance rule outputs a unique classification number. After the data is identified, the health management suggestion corresponding to the predicted classification number can be used as the target health management suggestion.

[0040] In the process of updating the parameters of deep learning algorithms in the prior art, there are technical problems such as being prone to falling into local optimality and slow updating speed, which results in poor final data classification effect and inability to provide accurate health guidance suggestions. Therefore, an embodiment of the present invention provides an algorithm for updating the parameters of the initial rules.

[0041] like Figure 2As shown, the parameters of the initial rule are updated by using the historical user physical data with an associated relationship and the unique classification number corresponding to the historical user physical data, including: S201, randomly initializing the parameters of the initial rule, and encoding the parameters of the initial rule using a vector encoding method to obtain an encoded individual; S202, repeatedly obtaining the coded individuals multiple times to obtain multiple different coded individuals; S203, after applying the parameters contained in the coded individual to the initial rule, taking the historical user body data as the input of the initial rule and taking the unique classification number corresponding to the historical user body data as the expected output, obtaining the fitness value corresponding to each coded individual; S204, based on the fitness value corresponding to each coding individual, determining an adaptive update step length, and searching the vicinity of the coding individual according to the adaptive update step length to determine the coding individual after the small range search; S205, based on the fitness value corresponding to each coding individual, determine the optimal coding individual, and determine the position information exchange individual according to the optimal coding individual, and use the position information exchange individual to further search the coding individuals after the small range search to determine the coding individuals after the extended search; S205, randomly matching the coded individual after the extended search with another coded individual after the extended search to obtain matching coded individuals corresponding to each coded individual after the extended search, and performing unknown area search between individuals on the coded individuals after the extended search according to the matching coded individuals to obtain coded individuals after the unknown area search; S206, determining the historical optimal state corresponding to each coding individual after the unknown area search, and performing an irregular search on the coding individual after the unknown area search according to the historical optimal state to determine the coding individual after the irregular search; S207, determining whether the current update number is greater than the preset maximum update number, if so, re-determining the optimal coding individual according to the coding individuals after the irregular search, otherwise returning to the step of obtaining the fitness value corresponding to each coding individual.

[0042] In the embodiment of the present invention, based on the fitness value corresponding to each coding individual, an adaptive update step is determined, and a nearby area of ​​the coding individual is searched according to the adaptive update step to determine the coding individual after the small range search, including: Based on the fitness value corresponding to each coded individual, the fitness ratio corresponding to the coded individual is determined as: ;in, Indicates i The fitness ratio corresponding to each encoded individual is Indicates i The fitness corresponding to each encoding individual, represents the maximum fitness corresponding to all encoded individuals, represents the minimum fitness corresponding to all encoded individuals, Represents a preset constant term; The fitness ratio corresponding to the encoded individual is used as the power of the power function, and combined with the update step size in the previous update process, the update step size in the current update process is adaptively determined, and the adaptive update step size is obtained as follows: ;in, represents the attenuation coefficient, and Greater than 1; Indicates the first i The update step size corresponding to the encoded individuals is Indicates the first i The update step size corresponding to the encoded individuals; Based on the adaptive update step size, different amounts of search are performed for the parameters of each dimension in the coding individual to achieve the search in the nearby area, and the coding individual obtained after the small range search is: ;in, Indicates the first i Coded individuals The d Dimension parameters; Indicates the first i Coded individuals The d Dimension parameter, that is, the number of encoded individuals after a small range search The d Dimension parameters; Represents a random number between (-1,1), Indicates d The difference between the upper threshold value and the lower threshold value of the dimension parameter; i =1,2,…,H, where H represents the total number of coded individuals. d =1,2,…,D, where D represents the total dimension of parameters in the encoding individual, that is, the total dimension of parameters of the deep learning algorithm.

[0043] The embodiment of the present invention forms an adaptive small-range update step based on the fitness corresponding to each coding individual, which can effectively realize the search of the coding individual within its neighborhood, so as to search for a better position near each coding individual, thereby realizing a preliminary coarse search.

[0044] In the embodiment of the present invention, based on the fitness value corresponding to each coding individual, the optimal coding individual is determined, and the position information exchange individual is determined according to the optimal coding individual, and the coding individual after the small range search is further searched using the position information exchange individual to determine the coding individual after the extended search, including: Based on the fitness value corresponding to each encoding individual, determine the optimal encoding individual ; For any coded individual after a small range search, randomly match the coded individual with the first random coded individual and the second randomly coded individual ; After exchanging information between the first random coding individual and the second random coding individual, an adaptive update step size is used for control, and combined with the position of the optimal coding individual, the position information exchange individual is obtained as ; After using Gaussian distribution to perturb the position information exchange individuals, the first search item corresponding to the coded individuals after a small range search is ;in, It represents a random number generated by a standard Gaussian distribution with a mean of 0 and a variance of 1. represents the update coefficient; The first search item is used to perform an extended search on the coded individuals after the narrow range search, and the coded individuals after the extended search are obtained as follows: ;in, Indicates the first search term d Dimension parameters, Indicates the first m Coded individuals after a small search The d Dimension parameters, Represents the coded individual after the extended search The d Dimension parameters, m =1,2,…,H.

[0045] By learning information from different coding individuals and the optimal coding individual, the embodiments of the present invention can not only enable the coding individuals to start from a better position, but also explore more unknown areas, improve the algorithm's traversal of high-dimensional solution space, and thus improve the performance of the algorithm in searching for a better solution.

[0046] In an embodiment of the present invention, a coding individual after an extended search is randomly matched with another coding individual after an extended search to obtain a matching coding individual corresponding to each coding individual after the extended search, and an unknown region search is performed between individuals on the coding individuals after the extended search according to the matching coding individuals to obtain the coding individuals after the unknown region search, including: Randomly match the coded individual after the extended search with another coded individual after the extended search, and obtain the matching coded individual corresponding to each coded individual after the extended search. ;in, Indicates j Encoded individuals after extended search The corresponding matching coded individuals, j =1,2,…,H; The coded individual after the extended search is exchanged with the corresponding matching coded individual, and the information exchanged individual is obtained as ; Based on the current update times, the information exchange individuals are adaptively searched and controlled using normal distribution, and the second search item corresponding to the coded individuals after the extended search is obtained as follows: ;in, Indicates the first d Dimension parameters, Indicates the first d Dimension parameters, It indicates that the attenuation coefficient is randomly generated according to the normal distribution N(0,1). Represents a random number between (0,1), Indicates d The upper threshold of the dimension parameter, Indicates d The lower threshold of the dimension parameter, T represents the preset maximum number of updates; The information exchange individual and the second search item corresponding to the coded individual are added together to determine the coded individual after the unknown area search: = + ;in, Represents the coded individual after the unknown area search The d Dimension parameters.

[0047] The embodiment of the present invention randomly matches a coding individual after an extended search with another coding individual after an extended search, and exchanges information to varying degrees, thereby further exploring the unknown area between the two coding individuals, improving the traversal of the algorithm in a high-dimensional solution space, and improving the performance of searching for the global optimal position.

[0048] In an embodiment of the present invention, determining the historical optimal state corresponding to each coding individual after the unknown area search, and performing an irregular search on the coding individual after the unknown area search according to the historical optimal state, and determining the coding individual after the irregular search, includes: Determine the historical optimal state of the encoded individual after searching each unknown area ;in, It represents the historical optimal state of the coded individual after the nth unknown area search. n =1,2,…,H; For any coded individual after searching in an unknown area, a Lvey flight step length is generated for it through the Lvey flight strategy. ; The Lvey flight step length is rounded to obtain the irregular information fusion quantity , and according to the number of irregular information fusions, multiple irregular decision individuals are paired for the coding individual; where INT represents the rounding function; According to the historical optimal states corresponding to multiple irregular decision individuals and coding individuals, irregular search is performed on the coding individuals to determine the irregular search individuals as follows: ;in, Indicates k The first of the irregular decision-making individuals d Dimension parameters, represents the first coded individual after the nth unknown area search d Dimension parameters, represents the d-th dimension parameter in the historical optimal state corresponding to the coding individual after the n-th unknown region search, Represents a random number between (-2, 2).

[0049] Determine whether the fitness value of the irregular search individual is greater than the fitness value corresponding to the coded individual after the unknown area search. If so, use the irregular search individual to update the coded individual after the unknown area search to obtain the coded individual after the irregular search. Otherwise, directly use the original coded individual after the unknown area search as the coded individual after the irregular search.

[0050] The present invention uses information fusion to perform irregular search, which can effectively use population information to realize reverse search, so that the coded individuals can search in areas far away from the population, thereby effectively improving the global search capability of the algorithm and avoiding the possibility of the algorithm falling into local optimality. At the same time, the greedy principle is used for control, which will not affect the convergence of the algorithm and can also improve the global search capability of the algorithm.

[0051] Optionally, the embodiment of the present invention may also perform out-of-bounds processing on the coding individual after each coding individual is changed, thereby ensuring that the algorithm parameters are within a valid range.

[0052] It should be understood by those skilled in the art that the embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes. The solutions in the embodiments of the present invention may be implemented in various computer languages, for example, object-oriented programming language Java and interpreted scripting language JavaScript, etc.

[0053] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks 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 generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0054] These computer program instructions may also be stored in a computer-readable memory capable of directing 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 an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0055] 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, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0056] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0057] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A health management platform based on deep learning, characterized in that: include: Physician registration module, physician data interaction module, deep learning module, user data interaction module and health guidance module; The physician registration module is used to obtain the registration request transmitted by the physician, and after reviewing the registration request, generate the physician account and password; The physician data interaction module is used to verify the physician account and password entered by the physician, and obtain the historical user physical data and corresponding health management suggestions entered by the physician after the verification is passed; The deep learning module is used to use a deep learning algorithm to perform data learning on the historical user physical data input by the physician and the corresponding health management suggestions to obtain health guidance rules; The user data interaction module is used to obtain the current user body data input by the user through human-computer interaction, and provide the current user body data to the health guidance module; The health guidance module is used to call the health guidance rules to process the current user's physical data, obtain target health management suggestions, and present the target health management suggestions to the user through a visual interface.

2. The deep learning-based health management platform according to claim 1, characterized in that: Obtain the registration request transmitted by the physician, and after reviewing the registration request, generate the physician account and password, including: Obtaining a registration request transmitted by a physician; wherein the registration request includes an ID number, a mobile phone number, a portrait image, an ID image, and a physician's certificate; Performing joint identity authentication using a portrait image, an ID card image, and a physician's certificate to obtain a first authentication result; wherein the first authentication result includes authentication success or authentication failure; Conducting an online query on the qualification certificate number on the physician certificate to determine a second verification result corresponding to the physician certificate; wherein the second verification result includes verification success or verification failure; Performing unified verification of the ID card number and the ID card image to determine a third verification result corresponding to the ID card; wherein the third verification result includes verification success or verification failure; When the first verification result, the second verification result and the third verification result are all successful, a unique physician account and password are randomly generated.

3. The deep learning-based health management platform according to claim 2, characterized in that: Historical user physical data includes: height, weight, gender, age, body mass index, waist circumference, waist-to-hip ratio, body fat percentage, and physical function assessment data.

4. The deep learning-based health management platform according to claim 3, characterized in that: After obtaining the historical user physical data and corresponding health management suggestions entered by the physician, it also includes: Determine the currently input health management suggestion, extract entities or keywords of the health management suggestion, and perform text similarity matching of the entities or keywords in the existing health management suggestions to obtain the target health management suggestion; Determine whether the text similarity between the currently input health management suggestion and the target health management suggestion is greater than a preset threshold, if so, determine that the currently input health management suggestion is a possible duplicate suggestion, otherwise determine that the currently input health management suggestion is a non-duplicate suggestion; When the currently input health management suggestion is a possible duplicate suggestion, the currently input health management suggestion is randomly transmitted to another physician for secondary judgment to obtain a secondary judgment result; wherein the secondary judgment result includes whether the currently input health management suggestion is a duplicate suggestion or a non-duplicate suggestion; When the secondary judgment result includes that the currently input health management suggestion is a duplicate suggestion, the health management suggestion corresponding to the historical user physical data input by the currently input physician is replaced with the target health management suggestion; When the secondary judgment result includes that the currently input health management suggestion is not a duplicate suggestion, the health management suggestion corresponding to the historical user physical data currently input by the physician is stored in the database; When the currently input health management suggestion is not a duplicate suggestion, the currently input health management is stored in the database.

5. The deep learning-based health management platform according to claim 4, characterized in that: A deep learning algorithm is used to learn the historical user physical data and corresponding health management suggestions input by the physician to obtain health guidance rules, including: Constructing a corresponding unique classification number for each health management suggestion, and associating the unique classification number with the historical user physical data input by the physician; A deep learning algorithm is used to construct the initial rules, and the parameters of the initial rules are updated through the historical user physical data with association and the unique classification number corresponding to the historical user physical data, so that the initial rules are adjusted to the data relationship between the historical user physical data and the unique classification number to obtain health guidance rules.

6. The deep learning-based health management platform according to claim 5, characterized in that: The parameters of the initial rule are updated by using the historical user physical data with an associated relationship and the unique classification number corresponding to the historical user physical data, including: The parameters of the initial rule are randomly initialized, and the parameters of the initial rule are encoded using a vector encoding method to obtain an encoded individual; Repeat the process of obtaining the coded individuals multiple times to obtain multiple different coded individuals; After applying the parameters contained in the coded individual to the initial rule, the historical user body data is used as the input of the initial rule, and the unique classification number corresponding to the historical user body data is used as the expected output to obtain the fitness value corresponding to each coded individual; Based on the fitness value corresponding to each coding individual, an adaptive update step is determined, and a nearby area of ​​the coding individual is searched according to the adaptive update step to determine the coding individual after a small range search; Based on the fitness value corresponding to each coding individual, the optimal coding individual is determined, and the position information exchange individual is determined based on the optimal coding individual. The position information exchange individual is used to further search the coding individuals after the small range search to determine the coding individuals after the extended search; Randomly matching the coded individual after the extended search with another coded individual after the extended search to obtain the matching coded individual corresponding to each coded individual after the extended search, and performing an unknown area search between the coded individuals after the extended search according to the matching coded individuals to obtain the coded individuals after the unknown area search; Determine the historical optimal state corresponding to each coding individual after the unknown area search, and perform an irregular search on the coding individual after the unknown area search according to the historical optimal state to determine the coding individual after the irregular search; Determine whether the current update number is greater than the preset maximum update number. If so, re-determine the optimal coding individual based on the coding individuals after the irregular search. Otherwise, return to the step of obtaining the fitness value corresponding to each coding individual.

7. The deep learning-based health management platform according to claim 6, characterized in that: Based on the fitness value corresponding to each coding individual, an adaptive update step is determined, and a nearby area of ​​the coding individual is searched according to the adaptive update step to determine the coding individual after a small range search, including: Based on the fitness value corresponding to each coded individual, determine the fitness ratio corresponding to the coded individual; Taking the fitness ratio corresponding to the encoded individuals as the power of the power function, and combining the update step in the previous update process, adaptively determining the update step in the current update process, and obtaining an adaptive update step; Based on the adaptive update step size, different amounts of search are performed for the parameters of each dimension in the coding individual to achieve a search in a nearby area and obtain the coding individual after a small-range search.

8. The deep learning-based health management platform according to claim 7, characterized in that: Based on the fitness value corresponding to each coding individual, the optimal coding individual is determined, and the position information exchange individual is determined based on the optimal coding individual. The position information exchange individual is used to further search the coding individuals after the small range search to determine the coding individuals after the extended search, including: Based on the fitness value corresponding to each coding individual, the optimal coding individual is determined; For any coded individual after searching in a small range, the coded individual is randomly matched with a first random coded individual and a second random coded individual; After exchanging information between the first random coding individual and the second random coding individual, an adaptive update step size is used for control, and the position information exchange individual is obtained in combination with the position of the optimal coding individual; After using Gaussian distribution to perturb the position information exchange individuals, the first search item corresponding to the coded individuals after a small range search is obtained; The first search item is used to perform an extended search on the coded individuals after the narrow range search to obtain coded individuals after the extended search.

9. The deep learning-based health management platform according to claim 8, characterized in that: Randomly matching the coded individual after the extended search with another coded individual after the extended search to obtain matching coded individuals corresponding to each coded individual after the extended search, and performing unknown area search between individuals on the coded individuals after the extended search according to the matching coded individuals to obtain coded individuals after the unknown area search, including: Randomly match the coded individual after the extended search with another coded individual after the extended search, and obtain the matching coded individual corresponding to each coded individual after the extended search; Performing information exchange between the coded individual after the extended search and the corresponding matching coded individual to obtain an information exchange individual; Based on the current update number, the information exchange individuals are adaptively searched and controlled using normal distribution to obtain a second search item corresponding to the coding individual after the extended search; The information exchange individual is added to the second search item corresponding to the coding individual to determine the coding individual after the unknown area search.

10. The deep learning-based health management platform according to claim 9, characterized in that: Determining the historical optimal state corresponding to each coding individual after the unknown area search, and performing an irregular search on the coding individual after the unknown area search according to the historical optimal state, and determining the coding individual after the irregular search, including: Determine the historical optimal state corresponding to the coding individual after searching each unknown area; For any coded individual after searching in an unknown area, a Lvey flight step is generated for it through the Lvey flight strategy; The Lvey flight step length is rounded to obtain the irregular information fusion quantity, and according to the irregular information fusion quantity, a plurality of irregular decision individuals are paired for the coding individual; According to the historical optimal states corresponding to the plurality of irregular decision individuals and the coding individuals, the coding individuals are irregularly searched to determine the irregular search individuals; Determine whether the fitness value of the irregular search individual is greater than the fitness value corresponding to the coded individual after the unknown area search. If so, use the irregular search individual to update the coded individual after the unknown area search to obtain the coded individual after the irregular search. Otherwise, directly use the original coded individual after the unknown area search as the coded individual after the irregular search.

Images