State detection processing method and device
By performing feature encoding and conversion of user data and combining dimensionality reduction processing, the shortcomings of the self-attention mechanism in information mining are solved, and the accuracy and effectiveness of user status detection are improved.
Patent Information
- Application Number
- CN202411894554.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-20
- Publication Date
- 2025-04-11
AI Technical Summary
When using artificial intelligence technology to process data, the information mining of the self-attention mechanism analysis of the data to be processed is insufficient, resulting in poor processing results.
Assisted user features are obtained by feature encoding of user data, and feature extraction is performed based on these features, feature conversion and information mining are performed for user features combined with auxiliary conversion features, and then the converted features are dimensionally reduced, the target user features are calculated and input into the user status detection model for detection.
The detection effect of user status detection is improved. By assisting the use of auxiliary features and transforming features, the user features are more accurate and comprehensive, and the accuracy and effectiveness of status detection are improved.
Smart Images
Figure CN119807781A_ABST
Abstract
Description
Technical Field
[0001] This document relates to the field of data processing, and in particular to a state detection processing method and device. Background Art
[0002] With the continuous advancement of digital construction, artificial intelligence technology is constantly combined with multiple fields; artificial intelligence technology can be used to efficiently and conveniently process data in multiple fields, such as classification, prediction and detection related data processing; in the process of using artificial intelligence technology for data processing, in order to improve the processing effect of artificial intelligence, the data to be processed is feature extracted and then further processed using artificial intelligence technology; At present, the self-attention mechanism can be used to analyze the important features in the data to be processed and increase the influence of the important features on the reasoning results, thereby improving the processing effect of data processing using artificial intelligence technology. However, analyzing the data to be processed through the self-attention mechanism, to a certain extent, does not sufficiently mine the information of the data to be processed, which leads to poor processing effect of data processing using artificial intelligence technology. Summary of the invention
[0003] An objective of an embodiment of the present specification is to provide a state detection processing method and device to improve the effectiveness of user state detection.
[0004] To solve the above technical problems, an embodiment of this specification is implemented as follows: In a first aspect, an embodiment of the present specification provides a state detection processing method, including: Performing feature encoding on the user data to obtain auxiliary user features, and performing feature extraction on the user data according to the auxiliary user features to obtain user features; Performing feature conversion on the user feature according to at least one auxiliary conversion feature to obtain a conversion feature; Performing dimensionality reduction processing on the conversion features, and calculating target user features based on the user data and the dimensionality reduction features obtained by the dimensionality reduction processing; The target user characteristics are input into a user state detection model to perform user state detection to obtain a detection result.
[0005] The state detection processing method provided in this embodiment, in the state detection processing process, firstly, feature encoding is performed on user data to obtain auxiliary user features, and then feature extraction is performed on the user data according to the auxiliary user features to obtain user features, so that the feature extraction of user data is affected by the auxiliary user features, so that the obtained user features are more accurate; after obtaining the user features, feature conversion is performed on the user features according to at least one auxiliary conversion feature to obtain conversion features, so that information mining is performed on the user features in combination with the auxiliary conversion features to further improve the accuracy of the obtained conversion features, on this basis, dimension reduction processing is performed on the conversion features, and target user features are calculated according to the user data and the dimension reduction features obtained by the dimension reduction processing, and then the target user features are input into the user state detection model to perform user state detection and obtain detection results, so that the feature extraction of user data is first assisted by the auxiliary user features to improve the accuracy of the obtained user features, and then information mining is performed on the user features in combination with the auxiliary conversion features to further improve the accuracy of the obtained conversion features, and the reduced dimension features are updated by introducing user data to improve the representation degree of the target user features to the user, thereby improving the detection effect of user state detection.
[0006] In a second aspect, another embodiment of the present specification provides a state detection processing device, including: A feature extraction module, used to perform feature encoding on user data to obtain auxiliary user features, and perform feature extraction on the user data according to the auxiliary user features to obtain user features; A feature conversion module, used to perform feature conversion on the user feature according to at least one auxiliary conversion feature to obtain a conversion feature; A dimension reduction processing module, used to perform dimension reduction processing on the conversion features, and calculate target user features based on the user data and the dimension reduction features obtained by the dimension reduction processing; The state detection module is used to input the target user characteristics into the user state detection model to perform user state detection and obtain a detection result.
[0007] In the third aspect, another embodiment of the present specification provides a state detection processing device, comprising: a memory, a processor, and computer executable instructions stored on the memory and executable on the processor, wherein the computer executable instructions, when executed by the processor, implement the steps of the state detection processing method as described in the first aspect above.
[0008] In a fourth aspect, another embodiment of the present specification provides a computer-readable storage medium, wherein the computer-readable storage medium is used to store computer-executable instructions, and when the computer-executable instructions are executed by a processor, the steps of the state detection processing method described in the first aspect above are implemented.
[0009] In a fifth aspect, another embodiment of the present specification provides a computer program product, wherein the computer program product includes a state detection processing program, and the state detection processing program is executed by a processor to implement the steps of the state detection processing method as described in the first aspect above. BRIEF DESCRIPTION OF THE DRAWINGS
[0010] In order to more clearly illustrate the technical solutions in one or more embodiments of the present specification, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0011] Figure 1 A state detection process flow chart provided for one embodiment of this specification; Figure 2 A feature extraction process flow chart provided for one embodiment of this specification; Figure 3 A feature conversion process flow chart provided for one embodiment of this specification; Figure 4 A schematic diagram of a feature conversion process provided by an embodiment of this specification; Figure 5 A process flow chart of a target user feature calculation process provided by one embodiment of this specification; Figure 6 A schematic diagram of a target user feature calculation process provided by an embodiment of this specification; Figure 7 A processing flow chart of a status detection processing method applied to a user health status detection scenario provided by an embodiment of this specification; Figure 8 A schematic diagram of a state detection processing device provided in one embodiment of this specification; Fig. 9 A schematic diagram of the structure of a state detection processing device provided in one embodiment of this specification. DETAILED DESCRIPTION
[0012] In order to enable those skilled in the art to better understand the technical solutions in one or more embodiments of this specification, the following will be combined with the drawings in one or more embodiments of this specification to clearly and completely describe the technical solutions in one or more embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on one or more embodiments of this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this document.
[0013] This specification provides an embodiment of a state detection processing method: Reference Figure 1 , which shows a processing flow chart of a state detection processing method provided in this embodiment. The state detection processing method provided in this embodiment specifically includes the following steps S102 to S108.
[0014] Step S102: feature encoding is performed on the user data to obtain auxiliary user features, and feature extraction is performed on the user data according to the auxiliary user features to obtain user features.
[0015] The user data in this embodiment may be relevant data of at least one dimension of the user required in the process of user status detection; for example, basic user data such as the user's age, gender, weight and / or address, or health indicator data such as blood pressure and / or blood lipids, or user activity data such as income status, number of daily activities and / or number of service participation. It should be noted that the user data may include one or more of the above data, and may also include other relevant data of the user according to the actual configuration, which is not limited in this embodiment. It should also be noted that the user status detection in this embodiment may be a user health status detection, or a service participation status detection of the user's participation in the service, which is not limited in this embodiment.
[0016] In specific implementation, since user data may include not only numerical values but also text to a certain extent, the processing effect and processing efficiency based on text processing are to a certain extent lower than numerical values. Based on this, after obtaining the user data, this embodiment can perform numerical conversion on the user data to obtain the user data represented by numerical values; in this embodiment, the user's relevant data before the numerical conversion can be represented by the user parameters, and the user data can represent the user's relevant data represented by numerical values. Optionally, the user data in this embodiment is obtained by converting the user parameters into numerical values; that is, the user data represents the user's relevant data in the form of numerical values.
[0017] In order to improve the accuracy and effectiveness of the obtained user data, in an optional implementation provided by this embodiment, after obtaining the user parameters of the user in at least one dimension, the user parameters of each dimension are numerically converted according to the user parameter categories of each dimension to obtain the user data of each dimension.
[0018] In an optional implementation manner provided by this embodiment, for a user parameter of any dimension in at least one dimension, numerical conversion is performed in the following manner: Detect whether the user parameter type of any dimension is a numeric type; If yes, determining the user parameter of the user in the any dimension as the user data of the user in the any dimension; If not, cluster the user parameters of any dimension to obtain parameter distribution categories, number the parameter distribution categories, and use the category numbers corresponding to the user parameters of the user in any dimension as the user data of the user in any dimension.
[0019] Specifically, if the user parameter type is a numerical type, the user parameter is determined as user data; if the user parameter type is a text type, the user parameter is converted into a numerical value to obtain the user data; that is, the user parameter of the numerical type is determined as the user data under the corresponding dimension, the user parameter of the text type is converted into a numerical value, and the numerical value obtained by the numerical conversion is determined as the user data under the corresponding dimension.
[0020] For example, at least one dimension includes user age and user gender; where user age is 64, which is a numeric type, and user gender is male, which is a text type, then the value 64 is directly used as the user data under the dimension of user age, and the user parameter male under the dimension of user gender is converted into a numeric value to obtain the category number as the user data under the dimension of user gender.
[0021] Furthermore, the numerical conversion of text-type user parameters can be performed through clustering numbering. In the process of performing numerical conversion of text-type dimensional user parameters through clustering numbering to obtain user data, the user parameters of the dimension can first be clustered to obtain parameter distribution categories, and then the parameter distribution categories can be numbered, and finally the category number corresponding to the user parameters of the user under the dimension is used as the user data of the user under the dimension.
[0022] Continuing with the above example, clustering is performed on the user parameters under the dimension of user gender, and the parameter distribution categories under the dimension of user gender are obtained, including male and female categories. The male and female categories are then numbered, with the male category numbered 1 and the female category numbered 2. In the case that the user parameter under the dimension of user gender is male, 1 is determined to be the user data under the dimension of user gender.
[0023] It should be noted that the state detection in this embodiment can be a multiple process. In order to improve the efficiency of numerical conversion in the state detection process and avoid repeated clustering and numbering processing of user parameters of the same non-numeric type, which results in long processing time and waste of resources, in this embodiment, when it is detected that the user parameter type of any dimension of the user is not a numeric type, the category number set corresponding to the any dimension can also be read, and the category number corresponding to the user parameter of the any dimension can be read from the category number set as the user data of the user under the any dimension.
[0024] After obtaining the user data of the user, feature extraction can be performed on the user data to obtain user features. In specific implementation, in order to avoid directly inputting the user data into the feature extraction model for feature extraction, which may result in a loose connection between the extracted user features, in this embodiment, after obtaining the user data, feature encoding is first performed on the user data to obtain auxiliary user features, and then feature extraction is performed on the user data based on the auxiliary user features to obtain user features.
[0025] During specific implementation, in order to better mine the fusion information of user data and avoid bringing a heavy burden to the server, in the process of feature encoding the user data to obtain auxiliary user features, the user data can be input into a single-layer neural network for feature extraction to obtain auxiliary user features; further, in order to make the connection between the obtained user features closer and thereby improve the accuracy of subsequent user state detection, the auxiliary user features and user data can be input into the embedding layer (embedding layer) for feature extraction to obtain user features.
[0026] Specifically, during the feature extraction process, the embedding layer may multiply the user data, the auxiliary user features, and the preset weights to obtain the user features.
[0027] For example, the embedding layer can be calculated as follows: ;in, Indicates the nth user data, represents the mth dimension of the auxiliary user matrix (auxiliary user features); Represents the value of the nth row and mth column of the user feature vector (user feature); Represents the weight vector of the nth user data; Represents the weight vector of the m-th dimension auxiliary user matrix. In this way, the calculated parameters are reduced and the extraction efficiency of the embedding layer for feature extraction is improved. When m is 16, For example, the amount of calculation can be reduced from 16n to .
[0028] In the calculation to obtain the user feature vector Finally, the feature vectors in the user feature matrix can be concatenated head to tail to obtain the concatenated user feature vector as the user feature. In this way, the auxiliary user features can be involved in the feature extraction of user data, and other sub-user features can affect the semantic conversion of a sub-user feature, thereby improving the closeness of the connection between sub-user features.
[0029] like Figure 2 The feature extraction processing flow chart shown in the figure, the process of performing feature encoding on user data to obtain auxiliary user features, and performing feature extraction on the user data according to the auxiliary user features to obtain user features, can be implemented by the following operations: Step S102-2, obtaining user parameters of at least one dimension.
[0030] Assume the number of dimensions is n.
[0031] Step S102 - 4 , performing numerical conversion on the user parameters of at least one dimension to obtain user data.
[0032] When the number of dimensions is n, the number of pieces of user data obtained is n.
[0033] Step S102-6: input the user data into a single-layer neural network for feature encoding to obtain auxiliary user features.
[0034] Assume that the auxiliary user features are 16-dimensional.
[0035] Step S102-8, performing embedding calculation on the auxiliary user features and user data to obtain a user feature matrix.
[0036] Wherein, when the number of users is n and the auxiliary user feature is a 16-dimensional feature vector, the user feature matrix has n rows and 16 columns. Each row of the user feature matrix is a feature vector after the user data of each dimension is converted.
[0037] Step S102-10, concatenate the feature vectors in the user feature matrix from beginning to end to obtain user features.
[0038] Among them, the user features are n*16 dimensions.
[0039] The above-mentioned feature encoding of user data to obtain auxiliary user features, and feature extraction of user data based on the auxiliary user features to obtain the process of user features are specifically described. After obtaining the user features in the above-mentioned manner, the internal attention mechanism can be used to perform feature conversion on the user features. The process of performing feature conversion on the user features using the internal attention mechanism is specifically described in the following step S104.
[0040] Step S104: performing feature conversion on the user feature according to at least one auxiliary conversion feature to obtain a conversion feature.
[0041] The auxiliary conversion feature may be a matrix of a preset number and a preset dimension.
[0042] Combination Figure 3 and Figure 4 ,The feature conversion process based on the internal attention mechanism can be implemented as follows.
[0043] Step S104-2, calculating a query matrix, a key matrix and a value matrix according to the user characteristics and an auxiliary conversion matrix of preset dimensions.
[0044] In the specific implementation, the user features are first calculated with different auxiliary transformation matrices of three preset dimensions to obtain a query matrix, a key matrix and a value matrix; specifically, the inner product of the user features and the first auxiliary transformation matrix can be calculated as the query matrix, the inner product of the user features and the second auxiliary transformation matrix can be calculated as the key matrix, and the inner product of the user features and the third auxiliary transformation matrix can be calculated as the value matrix.
[0045] Optionally, the preset dimension may be three dimensions.
[0046] Taking the inner product calculation of the first auxiliary conversion matrix and the user features as an example, in this embodiment, in the process of calculating the inner product of the user features and the first auxiliary conversion matrix, firstly, the n elements of the user features are multiplied by the elements of the first column of the first page of the three-dimensional matrix, respectively, to obtain n multiplier results, and then the n multiplier results are added to obtain 1 number as the first row and first column of the query matrix; repeat the above process until all m pages are calculated, and obtain a two-dimensional matrix of m rows and 1 columns as the query matrix; through the above method, each feature can interact, so that the query matrix obtained can obtain more feature interaction information. It should be noted that the key matrix and the value matrix are calculated in a similar manner, and this embodiment does not limit this.
[0047] For example, Figure 4 As shown, the first auxiliary transformation matrix is , the second auxiliary transformation matrix is , the third auxiliary transformation matrix is , calculate the inner products of the user features and the first auxiliary conversion matrix, the second auxiliary conversion matrix and the third auxiliary conversion matrix respectively, and obtain the m*l query matrix, key matrix and value matrix.
[0048] Taking the inner product calculation of the first auxiliary conversion matrix and the user features as an example, in this embodiment, in the process of calculating the inner product of the user features and the first auxiliary conversion matrix, firstly, the n elements of the user features are multiplied by the elements of the first column of the first page of the three-dimensional matrix, respectively, to obtain n multiplier results, and then the n multiplier results are added to obtain 1 number as the first row and first column of the query matrix; repeat the above process until all m pages are calculated, and obtain a two-dimensional matrix of m rows and 1 columns as the query matrix; through the above method, each feature can interact, so that the query matrix obtained can obtain more feature interaction information. It should be noted that the key matrix and the value matrix are calculated in a similar manner, and this embodiment does not limit this.
[0049] Specific, for example, user characteristics ,in, are all constants; ; Taking the calculation of the query matrix as an example, let
[0050]
[0051] in, express Page k of express The j-th row vector of , since the length of the feature vector of user features is 1, are also constants; assuming middle , the other row vectors are all 0, then:
[0052] ( ) Next order middle , the other row vectors are all 0, then:
[0053] At this point, the query matrix is obtained .
[0054] Step S104-4: Calculate an intermediate matrix according to the query matrix and the key matrix.
[0055] After the query matrix, key matrix and value matrix are calculated according to the user characteristics and the three randomly generated three-dimensional auxiliary transformation matrices, the intermediate matrix is first calculated according to the query matrix and the key matrix. In order to make the query matrix and the key matrix exchange information, an optional implementation provided by this embodiment first calculates the inner product matrix of the query matrix and the key matrix, and then normalizes the inner product matrix to obtain the intermediate matrix. In the process of calculating the inner product matrix of the query matrix (m*l) and the key matrix (m*l), the key matrix can be transposed first, and the inner product calculation is performed on the query matrix and the transposed key matrix (l*m) to obtain the inner product matrix (m*m) of the query matrix and the key matrix; further, in an optional implementation provided by this embodiment, the inner product matrix is normalized, and in the process of obtaining the intermediate matrix, the sum of the values of each row is calculated according to at least one value of each row in the inner product matrix, and then the updated value of each position is calculated according to the sum of the values of each position and the values of the row to which it belongs, and the intermediate matrix is generated according to the updated values of each position.
[0056] In the process of specifically calculating the inner product matrix of the query matrix and the key matrix, the l elements of the first row of the query matrix can be multiplied with the l elements of the first row of the transposed key matrix respectively to obtain l multiplier results, and the l multiplier results are added to obtain a value as the first row and first column of the inner product matrix, and so on to obtain the inner product matrix; after obtaining the inner product matrix, the m numbers in the first row of the inner product matrix can be added to obtain the sum of the values of the first row, and the updated values of each position are obtained by dividing the values of each position in the first row by the sum of the values, and for any row in the m rows, normalization is performed in this way to finally obtain the intermediate matrix.
[0057] For example, Figure 4 As shown, after the query matrix and the key matrix are obtained by calculation, the inner product matrix of the query matrix and the transposed key matrix is calculated, and the inner product matrix is normalized to obtain an intermediate matrix.
[0058] Step S104-6: Calculate a conversion matrix as the conversion feature according to the intermediate matrix and the value matrix.
[0059] After the intermediate matrix is calculated, the conversion features can be obtained by calculating the inner product of the intermediate matrix and the value matrix, for example, Figure 4 As shown, by calculating the inner product of the m*m intermediate matrix and the m*l value matrix, an m*l conversion matrix is obtained as the conversion feature.
[0060] In the process of calculating the inner product of the intermediate matrix and the value matrix, the m elements of the 1st row of the intermediate matrix can be multiplied by the m elements of the 1st row of the value matrix respectively to obtain m multiplier results, and the m multiplier results are added to obtain 1 number as the value of the 1st row and 1st column of the transformation matrix; calculate in a similar way to obtain the m*l transformation feature.
[0061] Step S106: perform dimensionality reduction processing on the converted features, and calculate target user features based on the user data and the dimensionality reduction features obtained by the dimensionality reduction processing.
[0062] In specific implementation, after obtaining the conversion features, the conversion features are reduced in dimension, and the target features are calculated based on the user data and the reduced in dimension features obtained by the dimensionality reduction process. Figure 6 , the dimension reduction processing is performed on the conversion features, and the process of calculating the target user features based on the user data and the dimension reduction features obtained by the dimension reduction processing is specifically described, such as Figure 5 As shown, the calculation process of the target user characteristics can be implemented through the following operations.
[0063] Step S106-2, calculating an intermediate auxiliary matrix according to the conversion feature and the dimension reduction auxiliary vector, and using the transposed matrix of the intermediate auxiliary matrix as the intermediate auxiliary feature.
[0064] In the specific implementation process, the intermediate auxiliary matrix is first calculated according to the transformation features and the dimension reduction auxiliary vector, and then the transposed matrix of the intermediate auxiliary matrix is used as the intermediate auxiliary feature.
[0065] For example, Figure 6 As shown, the inner product of the m*l conversion feature and the column vector of length l is calculated to obtain the m*1 intermediate auxiliary feature, and the transposed matrix of the intermediate auxiliary feature is used as the 1*m intermediate auxiliary feature.
[0066] In the process of calculating the inner product of the conversion feature and the dimension reduction auxiliary vector, the l elements of the first row of the conversion feature can be multiplied with the l elements of the dimension reduction auxiliary matrix respectively to obtain l multiplication results, and then the l multiplication results are added to obtain a number and placed in the first dimension of the intermediate auxiliary matrix. The inner product calculation is completed according to a similar process to obtain the intermediate auxiliary matrix.
[0067] Step S106-4, calculating the reduced dimension feature according to the intermediate auxiliary feature and the reduced dimension auxiliary matrix.
[0068] In the specific implementation process, after the intermediate auxiliary features are obtained as described above, the dimension reduction features are calculated based on the intermediate auxiliary features and the dimension reduction auxiliary matrix.
[0069] For example, Figure 6 As shown, the inner product of the 1*m intermediate auxiliary feature and the m*n dimension reduction auxiliary matrix can be calculated to obtain a 1*n vector as the dimension reduction feature.
[0070] In the process of calculating the inner product of the intermediate auxiliary feature and the dimensionality reduction auxiliary matrix, the m elements of the intermediate auxiliary feature can be multiplied with the m elements of the first row of the dimensionality reduction auxiliary matrix respectively to obtain m multiplier results, and then the m multiplier results are added to obtain 1 number, and the number is used as the first dimension of the dimensionality reduction feature. The inner product calculation is completed according to a similar process to obtain the dimensionality reduction matrix as the dimensionality reduction feature.
[0071] Step S106-6: Calculate target user features based on the user data and the dimension reduction features obtained by the dimension reduction process.
[0072] For example, Figure 6 As shown, the 1*n reduction feature and the 1*n user data are added to obtain the 1*n target user feature, so that the target user feature also has the user data.
[0073] Step S108: input the target user characteristics into a user state detection model to perform user state detection to obtain a detection result.
[0074] The user state detection model may be a neural network.
[0075] During the specific implementation, the target user characteristics are input into the user status model to perform user status detection to obtain the detection result. Specifically, if the detection result is biased towards 0, it can be determined that the user is in good health; if the detection result is biased towards 1, it can be determined that there is a problem with the user's health status, and an abnormal health warning for the user can be sent to the service user.
[0076] For example, when the test result is (0, , indicating that the status is good and no attention is needed; when the test result is (0.25, , indicating that there is a slight problem with the status and attention is needed; when the test result is (0.5, , indicating that there is a problem with the status and it needs to be taken seriously; when the test result is (0.75, 1), it means that there is a serious problem with the status and it needs to be resolved immediately.
[0077] In summary, the state detection method provided in this embodiment first preprocesses the user data to obtain user features. In the process of preprocessing the user data, the user data is first input into a single-layer neural network for feature extraction to obtain auxiliary user features, and then embedding calculation is performed based on the user data and the auxiliary user features to obtain user features. In this way, other feature information can affect the semantic conversion of a certain feature, making the connection between features closer and the conversion more accurate. Secondly, the user features are transformed according to at least one auxiliary transformation feature to obtain transformation features; in this process, the reference transformation matrix is first calculated according to the user features and the three-dimensional auxiliary transformation matrix, and the three-dimensional auxiliary transformation matrix is used to ensure that the information inside the user features is mined in large quantities to obtain more feature information. Further, the reference transformation matrix is normalized to obtain the transformation matrix; in this way, the importance relationship between different feature combinations is effectively distinguished, and by distinguishing the importance between feature combinations, the user state detection model can more accurately understand the feature information; On this basis, the conversion features are processed with dimensionality reduction to avoid the gradient vanishing problem caused by too many network layers during the conversion process. Then, the target user features are calculated based on the user data and the dimensionality reduction features, so that the target user features have the original user data. Finally, the target user features are input into the user state detection model to perform user state detection and obtain the detection results. In this way, the accuracy and effectiveness of user state detection can be improved by inputting target user data with more feature information and original user data for user state detection.
[0078] The following combination Figure 7 , taking the application of the state detection processing method provided by this embodiment in the user health status detection scenario as an example, the state detection processing method provided by this embodiment is further explained. Figure 7 The status detection processing method applied to the user health status detection scenario specifically includes steps S702 to S714.
[0079] Step S702, obtaining user parameters of at least one dimension, and performing numerical conversion on the user parameters to obtain user data.
[0080] Step S704: input the user data into a single-layer neural network for feature encoding to obtain auxiliary user features.
[0081] Step S706, calculating user features based on the auxiliary user features and the user data.
[0082] Step S708, calculating the query matrix, key matrix and value matrix according to the user characteristics and the auxiliary conversion matrix.
[0083] Step S710, calculating the inner product matrix of the query matrix and the key matrix, normalizing the inner product matrix, and obtaining an intermediate matrix.
[0084] Step S712, calculate the conversion features according to the intermediate matrix and the value matrix, and input the conversion features into the residual network for dimensionality reduction processing to obtain the target user features.
[0085] Step S714: Input the target user characteristics into the user health status detection model to perform user health status detection to obtain the user health status detection result.
[0086] It should be noted that any one of steps S702 to S714 or a combination of any multiple steps can be selected from steps S102 to S108 to form a new implementation method according to the needs of implementation deployment; and any one or multiple technical features in the technical scheme composed of steps S702 to S714 can be selected from steps S102 to S108 to form a new implementation method according to the needs of actual deployment, or the technical features in one or more optional implementation methods provided by steps S102 to S108 can be selected to form a new implementation method, which will not be described one by one here.
[0087] Figure 8 A schematic diagram of a state detection processing device provided by an embodiment of the present invention is shown in FIG. Figure 8 As shown, the device comprises: The feature extraction module 802 is used to perform feature encoding on the user data to obtain auxiliary user features, and perform feature extraction on the user data according to the auxiliary user features to obtain user features; A feature conversion module 804 is used to perform feature conversion on the user feature according to at least one auxiliary conversion feature to obtain a conversion feature; A dimension reduction processing module 806 is used to perform dimension reduction processing on the conversion features, and calculate target user features based on the user data and the dimension reduction features obtained by the dimension reduction processing; The state detection module 808 is used to input the target user characteristics into the user state detection model to perform user state detection and obtain a detection result.
[0088] The state detection processing device provided in this embodiment, in the state detection processing process, firstly performs feature encoding on user data to obtain auxiliary user features, and then performs feature extraction on user data according to the auxiliary user features to obtain user features, thereby using the auxiliary user features to influence the feature extraction of user data, so that the obtained user features are more accurate; after obtaining the user features, the user features are feature converted according to at least one auxiliary conversion feature to obtain conversion features, thereby, information mining is performed on the user features in combination with the auxiliary conversion features to further improve the accuracy of the obtained conversion features, on this basis, dimension reduction processing is performed on the conversion features, and target user features are calculated according to the user data and the dimension reduction features obtained by the dimension reduction processing, and then the target user features are input into the user state detection model to perform user state detection and obtain detection results, thereby, firstly, the feature extraction of user data is assisted by the auxiliary user features to improve the accuracy of the obtained user features, and then information mining is performed on the user features in combination with the auxiliary conversion features to further improve the accuracy of the obtained conversion features, and the user data is introduced to update the dimension reduction features to improve the representation degree of the target user features to the user, thereby improving the detection effect of the user state detection.
[0089] The state detection processing device provided in one embodiment of this specification can implement each process in the aforementioned method embodiment and achieve the same functions and effects, which will not be repeated here.
[0090] Furthermore, an embodiment of the present specification also provides a state detection processing device, Fig. 9 A schematic diagram of a state detection processing device provided in one embodiment of this specification is shown in FIG. Fig. 9 As shown, the device includes: a memory 901, a processor 902, a bus 903 and a communication interface 904. The memory 901, the processor 902 and the communication interface 904 communicate through the bus 903, and the communication interface 904 may include an input and output interface, which includes but is not limited to a keyboard, a mouse, a display, a microphone, a loudspeaker, etc.
[0091] Fig. 9 In the embodiment, the memory 901 stores computer executable instructions that can be run on the processor 902. When the computer executable instructions are executed by the processor 902, the following process is implemented: Performing feature encoding on the user data to obtain auxiliary user features, and performing feature extraction on the user data according to the auxiliary user features to obtain user features; Performing feature conversion on the user feature according to at least one auxiliary conversion feature to obtain a conversion feature; Performing dimensionality reduction processing on the conversion features, and calculating target user features based on the user data and the dimensionality reduction features obtained by the dimensionality reduction processing; The target user characteristics are input into a user state detection model to perform user state detection to obtain a detection result.
[0092] The state detection processing device provided in this embodiment, through the cooperation of the memory 901, the processor 902, the bus 903 and the communication interface 904, firstly performs feature encoding on the user data to obtain auxiliary user features in the state detection processing process, and then performs feature extraction on the user data according to the auxiliary user features to obtain user features, thereby influencing the feature extraction of the user data with the help of the auxiliary user features, so that the obtained user features are more accurate; after obtaining the user features, the user features are subjected to feature conversion according to at least one auxiliary conversion feature to obtain conversion features, thereby, information mining is performed on the user features in combination with the auxiliary conversion features to further improve the accuracy of the obtained conversion features, on this basis, dimension reduction processing is performed on the conversion features, and target user features are calculated according to the user data and the dimension reduction features obtained by the dimension reduction processing, and then the target user features are input into the user state detection model to perform user state detection and obtain detection results, thereby, firstly, the feature extraction of the user data is assisted by the auxiliary user features to improve the accuracy of the obtained user features, and then information mining is performed on the user features in combination with the auxiliary conversion features to further improve the accuracy of the obtained conversion features, and the reduced dimension features are updated by introducing user data to improve the representation degree of the target user features to the user, thereby improving the detection effect of the user state detection.
[0093] The state detection processing device provided in one embodiment of this specification can implement each process in the aforementioned method embodiment and achieve the same functions and effects, which will not be repeated here.
[0094] Furthermore, another embodiment of the present specification provides a computer-readable storage medium, wherein the computer-readable storage medium is used to store computer-executable instructions, and when the computer-executable instructions are executed by a processor, the following process is implemented: Performing feature encoding on the user data to obtain auxiliary user features, and performing feature extraction on the user data according to the auxiliary user features to obtain user features; Performing feature conversion on the user feature according to at least one auxiliary conversion feature to obtain a conversion feature; Performing dimensionality reduction processing on the conversion features, and calculating target user features based on the user data and the dimensionality reduction features obtained by the dimensionality reduction processing; The target user characteristics are input into a user state detection model to perform user state detection to obtain a detection result.
[0095] The computer-readable storage medium provided in the present embodiment, in the process of state detection processing, firstly performs feature encoding on user data to obtain auxiliary user features, then performs feature extraction on user data according to the auxiliary user features to obtain user features, thereby using the auxiliary user features to influence the feature extraction of user data, so that the obtained user features are more accurate; after obtaining the user features, the user features are feature converted according to at least one auxiliary conversion feature to obtain conversion features, thereby, information mining is performed on the user features in combination with the auxiliary conversion features to further improve the accuracy of the obtained conversion features, on this basis, dimension reduction processing is performed on the conversion features, and target user features are calculated according to the user data and the dimension reduction features obtained by the dimension reduction processing, and then the target user features are input into the user state detection model to perform user state detection and obtain detection results, thereby, firstly, the feature extraction of user data is assisted by the auxiliary user features to improve the accuracy of the obtained user features, then information mining is performed on the user features in combination with the auxiliary conversion features to further improve the accuracy of the obtained conversion features, and the reduced dimension features are updated by introducing user data to improve the representation degree of the target user features to the user, thereby improving the detection effect of user state detection.
[0096] The computer-readable storage medium includes a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.
[0097] The computer-readable storage medium provided in one embodiment of this specification can implement each process in the aforementioned method embodiment and achieve the same functions and effects, which will not be repeated here.
[0098] Furthermore, another embodiment of the present specification also provides a computer program product, which implements the following process when executed by a processor: Performing feature encoding on the user data to obtain auxiliary user features, and performing feature extraction on the user data according to the auxiliary user features to obtain user features; Performing feature conversion on the user feature according to at least one auxiliary conversion feature to obtain a conversion feature; Performing dimensionality reduction processing on the conversion features, and calculating target user features based on the user data and the dimensionality reduction features obtained by the dimensionality reduction processing; The target user characteristics are input into a user state detection model to perform user state detection to obtain a detection result.
[0099] The computer program product provided by this embodiment, in the process of state detection processing, firstly performs feature encoding on user data to obtain auxiliary user features, then performs feature extraction on user data according to the auxiliary user features to obtain user features, thereby using the auxiliary user features to influence the feature extraction of user data, so that the obtained user features are more accurate; after obtaining the user features, the user features are subjected to feature conversion according to at least one auxiliary conversion feature to obtain conversion features, thereby, information mining is performed on the user features in combination with the auxiliary conversion features to further improve the accuracy of the obtained conversion features, on this basis, dimension reduction processing is performed on the conversion features, and target user features are calculated according to the user data and the dimension reduction features obtained by the dimension reduction processing, and then the target user features are input into the user state detection model to perform user state detection and obtain detection results, thereby, firstly, the feature extraction of user data is assisted by the auxiliary user features to improve the accuracy of the obtained user features, then information mining is performed on the user features in combination with the auxiliary conversion features to further improve the accuracy of the obtained conversion features, and the reduced dimension features are updated by introducing user data to improve the representation degree of the target user features to the user, thereby improving the detection effect of user state detection.
[0100] The computer program product provided in one embodiment of this specification can implement each process in the aforementioned method embodiment and achieve the same functions and effects, which will not be repeated here.
[0101] It will be appreciated by those skilled in the art that 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. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-readable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0102] 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.
[0103] 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.
[0104] 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.
[0105] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.
[0106] Memory may include non-permanent storage in a computer-readable storage medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable storage medium.
[0107] Computer-readable storage media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer-readable instructions, data structures, program modules or other data. Examples of computer-readable storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable storage media does not include temporary computer-readable media (transitory media), such as modulated data signals and carrier waves.
[0108] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.
[0109] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, the present application 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.
[0110] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.
Claims
1. A state detection processing method, characterized in that: The method comprises: Performing feature encoding on the user data to obtain auxiliary user features, and performing feature extraction on the user data according to the auxiliary user features to obtain user features; Performing feature conversion on the user feature according to at least one auxiliary conversion feature to obtain a conversion feature; Performing dimensionality reduction processing on the conversion features, and calculating target user features based on the user data and the dimensionality reduction features obtained by the dimensionality reduction processing; The target user characteristics are input into a user state detection model to perform user state detection to obtain a detection result.
2. The state detection processing method according to claim 1, characterized in that: Before the step of encoding the user data to obtain auxiliary user features and extracting features from the user data based on the auxiliary user features, the method further includes: Obtain user parameters of the user in at least one dimension; The user parameters of each dimension are numerically converted according to the user parameter categories of each dimension to obtain the user data of each dimension.
3. The state detection processing method according to claim 1, characterized in that: The step of performing feature conversion on the user feature according to at least one auxiliary conversion feature to obtain a conversion feature includes: Calculate a query matrix, a key matrix and a value matrix according to the user characteristics and an auxiliary conversion matrix of preset dimensions; Calculating an intermediate matrix according to the query matrix and the key matrix; A conversion matrix is calculated as the conversion feature according to the intermediate matrix and the value matrix.
4. The state detection processing method according to claim 3, characterized in that: The step of calculating an intermediate matrix according to the query matrix and the key matrix comprises: Calculating an inner product matrix of the query matrix and the key matrix; The inner product matrix is normalized to obtain the intermediate matrix.
5. The state detection processing method according to claim 4, characterized in that: The step of normalizing the inner product matrix to obtain the intermediate matrix includes: Calculating the sum of the values of each row in the inner product matrix according to at least one value of each row; The updated values of each position are calculated according to the value of each position and the value of the corresponding row, and the intermediate matrix is generated according to the updated values of each position.
6. The state detection processing method according to claim 1, characterized in that: The performing dimensionality reduction processing on the conversion feature includes: Calculate an intermediate auxiliary matrix according to the conversion feature and the dimension reduction auxiliary vector, and use the transposed matrix of the intermediate auxiliary matrix as the intermediate auxiliary feature; The dimension reduction feature is calculated according to the intermediate auxiliary feature and the dimension reduction auxiliary matrix.
7. A state detection processing device, characterized in that: The device comprises: A feature extraction module, used to perform feature encoding on user data to obtain auxiliary user features, and perform feature extraction on the user data according to the auxiliary user features to obtain user features; A feature conversion module, used to perform feature conversion on the user feature according to at least one auxiliary conversion feature to obtain a conversion feature; A dimension reduction processing module, used to perform dimension reduction processing on the conversion features, and calculate target user features based on the user data and the dimension reduction features obtained by the dimension reduction processing; The state detection module is used to input the target user characteristics into the user state detection model to perform user state detection and obtain a detection result.
8. A state detection processing device, characterized in that: The device includes a memory and a processor, wherein the memory stores computer executable instructions, and when the computer executable instructions are executed on the processor, the steps of the state detection processing method according to any one of claims 1 to 6 can be implemented.
9. A computer-readable storage medium having computer-executable instructions stored therein, characterized in that: When the computer executable instructions are executed by a processor, they can implement the state detection processing method described in any one of claims 1 to 6.
10. A computer program product, characterized in that The computer program product includes a data processing program, and when the data processing program is executed by a processor, the steps of the state detection processing method described in any one of claims 1 to 6 can be implemented.