A user feature processing method, device and medium
By automating the segmentation and filtering of user feature subsets, the problem of low efficiency in manually extracting training samples is solved, thereby improving sample quality and the accuracy of prediction models.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- ZHEJIANG MEIRI HUDONG NETWORK TECH CO LTD
- Filing Date
- 2026-04-13
- Publication Date
- 2026-07-10
AI Technical Summary
In existing technologies, the extraction of positive and negative training samples from user feature data relies on manual operation, which leads to low efficiency and is prone to labeling errors, affecting the accuracy of the prediction model.
The user feature set is automatically divided into an intermediate feature subset, a first feature subset, and a second feature subset. The target feature subset is then selected as a training sample based on the status label of the feature subset, thereby reducing the number of features to be analyzed and improving the efficiency of training sample extraction.
The system enables automated extraction of training samples, reduces manual intervention, improves sample extraction efficiency and quality, and ensures the accuracy of the prediction model.
Smart Images
Figure CN122020105B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of electronic digital data processing technology, and in particular to a method, apparatus and medium for processing user characteristics. Background Technology
[0002] In the field of user status prediction, such as in healthcare scenarios, it is typically necessary to extract positive and negative training samples from users' historical feature data to train the prediction model. The quality and efficiency of extracting these training samples directly determine the model's effectiveness and iteration speed. Traditional solutions rely heavily on manual extraction of positive and negative training samples, which presents several problems: manual sifting of samples from a large amount of historical user data is time-consuming and inefficient; furthermore, errors can occur during manual extraction, leading to incorrect labeling of samples, reducing training sample quality, and impacting the accuracy of the prediction model. Therefore, improving the efficiency of training sample extraction is a pressing issue. Summary of the Invention
[0003] The purpose of this invention is to provide a method, device, and medium for processing user features to improve the efficiency of training sample extraction.
[0004] According to a first aspect of the present invention, a method for processing user features is provided, the method comprising the following steps:
[0005] Obtain the target user feature set; the target user feature set includes feature sets of several users, each feature set includes several feature subsets, and each feature subset includes several feature data of a preset type with the same time attribute.
[0006] For any user's feature set, the feature subsets whose time attribute values are equal to, less than, and greater than the median value are respectively determined as the user's intermediate feature subset, first feature subset, and second feature subset.
[0007] If the status label of a user corresponding to any intermediate feature subset is the first preset label, then the target feature subset of the user is selected from the second feature subset of the user; otherwise, the target feature subset of the user is selected from the first feature subset of the user. The target feature subset of the user satisfies the following: the status label of the user corresponding to the feature subset whose time attribute value is less than the time attribute value of the target feature subset is the first preset label, and the status label of the user corresponding to the feature subset whose time attribute value is greater than the time attribute value of the target feature subset is the second preset label.
[0008] If the screening is successful, the target feature subset of the user is determined as the positive sample when training the target model; the target model is used to obtain the user's state label.
[0009] Furthermore, filtering the user's target feature subset from the user's second feature subset includes:
[0010] The second feature subset is sorted in ascending order of time attribute values to obtain the first feature subset sequence.
[0011] Using a first preset value as the step size, the judgment starts from the nth feature subset with the smallest time attribute value in the first feature subset sequence, where n is the first preset value.
[0012] If the status label of the user corresponding to the smallest nth feature subset is the second preset label, then continue to determine the status label of the user corresponding to the (n-1)th feature subset with the smallest time attribute value in the feature subset sequence.
[0013] If the status label of the user corresponding to the smallest (n-1)th feature subset is the first preset label, then the smallest (n-1)th feature subset is determined as the target feature subset; if the status label of the user corresponding to the smallest (n-1)th feature subset is the second preset label, then the judgment continues in the direction of decreasing time attribute value until the status label of the user corresponding to a certain feature subset is determined to be the first preset label, and the next feature subset of that feature subset is determined as the target feature subset.
[0014] Furthermore, the process of obtaining the first preset value includes:
[0015] Obtain the difference between the maximum and minimum values of the time attributes of the feature subset included in the second feature subset.
[0016] The average time interval of the time attribute values of adjacent feature subsets included in the second feature subset is obtained based on the difference and the number of feature subsets included in the second feature subset.
[0017] The feature subset density per unit time is obtained based on the average time interval.
[0018] A first preset value is obtained based on the feature subset density; the first preset value and the feature subset density are positively correlated.
[0019] Furthermore, filtering the user's target feature subset from the user's second feature subset also includes:
[0020] If the status label of the user corresponding to the smallest nth feature subset is the first preset label, then continue to determine the status label of the user corresponding to the 2nth feature subset with the smallest time attribute value in the feature subset sequence; if the status label of the user corresponding to the smallest 2nth feature subset is the first preset label, then continue to determine in the direction of increasing time attribute value.
[0021] Furthermore, the process of obtaining the target user feature set includes:
[0022] The initial user feature set is filtered according to preset filtering conditions to obtain the target user feature set. The initial user feature set includes the feature sets of several users. The preset filtering conditions include: the number of feature subsets included in the user's feature set is greater than or equal to the target preset number, and the difference between the maximum and minimum values of the time attributes of the feature subsets included in the user's feature set is greater than or equal to the target preset time interval.
[0023] Furthermore, the process of obtaining the target user feature set includes:
[0024] For any feature subset of any user, if the feature data of a certain preset type in the feature subset of the user is empty, then the feature data of the preset type in the feature subset of the user is obtained based on the feature data of the same preset type in the similar feature subset of the user. The similar feature subset of the user satisfies the following condition: the time attribute value is less than the time attribute value of the feature subset, and the difference between the time attribute value and the time attribute value of the feature subset is less than or equal to a preset time threshold.
[0025] Furthermore, the method also includes: identifying any subset of features whose time attribute value is less than the time attribute value of the target feature subset of the user as negative samples when training the target neural network model.
[0026] Furthermore, the preset type of feature data includes at least one of the following types of feature data: age, gender, exercise frequency, blood glucose, blood pressure, blood lipids, alcohol consumption frequency, smoking frequency, medication feature data, imaging examination results, family history feature data, and past medical history feature data.
[0027] According to a second aspect of the present invention, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the processing method of the user features described above.
[0028] According to a third aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for processing user features.
[0029] Compared with the prior art, the present invention has at least the following beneficial effects:
[0030] This invention, for any user's feature set in the target user feature set, determines the feature subset with the median time attribute value as the user's intermediate feature subset, the feature subset with the time attribute value less than the median value as the user's first feature subset, and the feature subset with the time attribute value greater than the median value as the user's second feature subset. If the user's status label corresponding to the intermediate feature subset is a first preset label, then the target feature subset for that user is selected from the user's second feature subset; otherwise, the target feature subset for that user is selected from the user's first feature subset. The target feature subset for each user is determined as the positive sample for training the target model. Compared to existing technologies, this invention not only automates the extraction of training samples, but also divides the user's feature set into a first feature subset and a second feature subset, selecting the target feature subset (positive sample) only from the first or second feature subset, thus reducing the number of features to be analyzed and improving the efficiency of training sample extraction. Attached Figure Description
[0031] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0032] Figure 1 A flowchart of a user feature processing method provided in Embodiment 1 of the present invention;
[0033] Figure 2 This is a flowchart of the steps for filtering a target feature subset of a user from a second feature subset provided in Embodiment 1 of the present invention;
[0034] Figure 3 This is a flowchart of the process for obtaining the first preset value provided in Embodiment 1 of the present invention. Detailed Implementation
[0035] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0036] Example 1:
[0037] According to this embodiment, as Figure 1 As shown, a method for processing user characteristics is provided, the method comprising the following steps:
[0038] S100, Obtain the target user feature set; the target user feature set includes feature sets of several users, each feature set includes several feature subsets, and each feature subset includes several feature data of a preset type with the same time attribute.
[0039] As a specific implementation, the process of obtaining the target user feature set includes: filtering the initial user feature set according to preset filtering conditions to obtain the target user feature set; the initial user feature set includes feature sets of several users, and the preset filtering conditions include: the number of feature subsets included in a user's feature set is greater than or equal to a target preset number, and the difference between the maximum and minimum values of the time attributes of the feature subsets included in a user's feature set is greater than or equal to a target preset time interval. This allows for the exclusion of users with insufficient numbers or short time spans, ensuring the quality and representativeness of users in the target user feature set. Optionally, the target preset number and target preset time interval are empirical values, for example, the target preset number is 2 or 3, and the target preset time interval is 1 year or half a year.
[0040] As a specific implementation, the process of obtaining the target user feature set includes: for any feature subset of any user, if the feature data of a certain preset type in the feature subset of the user is empty, then the feature data of the preset type in the feature subset of the user is obtained based on the feature data of the similar feature subsets of the user; the similar feature subsets of the user satisfy the following condition: the time attribute value is less than the time attribute value of the feature subset (denoted as the first condition), and the difference between the time attribute value and the time attribute value of the feature subset is less than or equal to a preset time threshold. Thus, missing values in the feature subset can be filled in, ensuring that the data can be used for subsequent analysis. Optionally, the preset time threshold is the time attribute value of the m-th feature subset with the largest time attribute value in the feature subset of the user that satisfies the first condition, where m is a preset number, for example, m is 1, 2, or 3, etc. It should be understood that the largest m-th is: arranged in descending order, in the m-th position.
[0041] As a specific implementation, obtaining the feature data of the preset type in the user's feature subset based on the feature data of the similar feature subsets of the user's feature subset includes: if there are feature subsets of the preset type with no missing feature data among m similar feature subsets, then the mean, minimum, or maximum value of the feature data of the preset type in these feature subsets is determined as the feature data of the preset type in the user's feature subset. Otherwise, m is updated to m+1, and the processing is repeated, or the feature data of the preset type in the user's feature subset is determined to be empty.
[0042] In this embodiment, the preset type of feature data is feature data related to the user's status. As a specific implementation, positive and negative stroke samples are extracted from the target user feature set. The preset type of feature data includes at least one of the following types: age, gender, exercise frequency, blood glucose, blood pressure, blood lipids, alcohol consumption frequency, smoking frequency, medication characteristics, imaging examination results, family history characteristics, and past medical history characteristics. Medication characteristics include medication name, dosage, and frequency; imaging examination results refer to medical imaging examination results, such as carotid artery stenosis examination results; family history characteristics refer to family history information, such as whether there is a family history of cardiovascular and cerebrovascular diseases; and past medical history characteristics refer to historical illness information, such as coronary heart disease and carotid artery stenosis.
[0043] S200, for any user's feature set, the feature subsets whose time attribute values are equal to, less than, and greater than the median value are respectively determined as the user's median feature subset, first feature subset, and second feature subset.
[0044] In this embodiment, for each user's feature set, after sorting by time attribute value, the feature subset corresponding to the median value is found as the benchmark, and the feature subset is divided into three parts: the first feature subset (time value is less than the median value), the second feature subset (time value is greater than the median value), and the median feature subset.
[0045] In one specific implementation, the median or average of the time attribute values of all feature subsets for any user is used as the median value. The relative magnitudes of the time attribute values of each feature subset are then compared to the median value to determine the subset. For example, if the time attribute is date and the median value is June 1, 2023, then the feature subset with time attribute values before June 1, 2023 is the first feature subset, and the data with time attribute values after June 1, 2023 is the second feature subset.
[0046] S300, if the status label of the user corresponding to the intermediate feature subset of any user is the first preset label, then the target feature subset of the user is selected from the second feature subset of the user; otherwise, the target feature subset of the user is selected from the first feature subset of the user; the target feature subset of the user satisfies the following: the status label of the user corresponding to the feature subset whose time attribute value is less than the time attribute value of the target feature subset is the first preset label, and the status label of the user corresponding to the feature subset whose time attribute value is greater than the time attribute value of the target feature subset is the second preset label.
[0047] In this embodiment, the search direction is determined based on the state labels of the intermediate feature subset: if the intermediate feature subset is a first preset label, then the search proceeds from the second feature subset to the critical point where the state transitions from the first preset label to the second preset label; otherwise, the search continues from the first feature subset. The target feature subset must satisfy the condition that all its preceding state labels are the first preset label, and all its subsequent state labels are the second preset label, i.e., a state inflection point. As a specific implementation, based on user feature extraction, positive and negative samples of user stroke are identified. The corresponding first preset label indicates no stroke has occurred, and the second preset label indicates a stroke has occurred.
[0048] As a preferred embodiment, filtering the user's target feature subset from the user's second feature subset includes, for example: Figure 2 As shown:
[0049] S310, Sort the second feature subset according to the time attribute value from smallest to largest to obtain the first feature subset sequence.
[0050] S320, with a first preset value as the step size, start judging from the nth feature subset with the smallest time attribute value in the first feature subset sequence, where n is the first preset value.
[0051] It should be understood that the smallest nth is the nth position when arranged in ascending order.
[0052] As a specific implementation method, the process of obtaining the first preset value includes, for example: Figure 3 As shown:
[0053] S321, obtain the difference between the maximum and minimum values of the time attributes of the feature subsets included in the second feature subset.
[0054] S322, based on the difference and the number of feature subsets included in the second feature subset, obtain the average time interval of the time attribute values of the adjacent feature subsets included in the second feature subset.
[0055] As a specific implementation, the average time interval is a, a=t / (q-1), where t is the difference between the maximum and minimum values of the time attributes of the feature subsets included in the second feature subset, and q is the number of feature subsets included in the second feature subset.
[0056] S323, obtain the feature subset density per unit time based on the average time interval.
[0057] As a specific implementation, the feature subset density per unit time is D, where D = T / a, and T is a preset unit time, which can be selected as 1 year or half a year, etc.
[0058] S324, Obtain a first preset value based on the feature subset density; the first preset value and the feature subset density are positively correlated.
[0059] In one specific implementation, the first preset value is n, where n = min(q, max(1, round(B×D / D0))), where the round function represents rounding to the nearest integer, B is the basic preset value, D0 is the reference density, min() is to take the minimum value, and max() is to take the maximum value. B and D0 are empirical values, for example, B is 2 or 3, and D0 is 2 or 4. Based on this specific implementation, the first preset value is adjusted according to the feature subset density, increasing the first preset value (i.e., increasing the step size) when the density is high and decreasing the first preset value (i.e., decreasing the step size) when the density is low, which helps improve the efficiency of subsequently obtaining the target feature subset.
[0060] S330, if the status label of the user corresponding to the smallest nth feature subset is the second preset label, then continue to determine the status label of the user corresponding to the (n-1)th feature subset with the smallest time attribute value in the feature subset sequence.
[0061] S340, if the status label of the user corresponding to the smallest (n-1)th feature subset is the first preset label, then the smallest nth feature subset is determined as the target feature subset; if the status label of the user corresponding to the smallest (n-1)th feature subset is the second preset label, then the judgment continues in the direction of decreasing time attribute value until the status label of the user corresponding to a certain feature subset is determined to be the first preset label, and the next feature subset of that feature subset is determined as the target feature subset.
[0062] In this embodiment, the process of filtering the target feature subset of the user from the second feature subset of the user further includes: if the status label of the user corresponding to the smallest nth feature subset is the first preset label, then continue to determine the status label of the user corresponding to the 2nth feature subset with the smallest time attribute value in the feature subset sequence; if the status label of the user corresponding to the smallest 2nth feature subset is the first preset label, then continue to determine along the direction of increasing time attribute value.
[0063] Based on S310-S340, when the user's status label corresponding to the intermediate feature subset is the first preset label, the user's status label is determined starting from the nth feature subset with the smallest time attribute value in the first feature subset sequence. If the label is the first preset label, the user jumps in the direction of increasing time (with the first preset value as the step size) for inspection; if the label is the second preset label, the user checks step by step in the direction of decreasing time until a turning point is found.
[0064] In this embodiment, the process of filtering the user's target feature subset from the user's first feature subset is similar to the process of filtering the user's target feature subset from the user's second feature subset described above. The differences include: sorting the first feature subset according to the time attribute value from smallest to largest to obtain the second feature subset sequence; using a second preset value p as the step size, starting from the p-th feature subset with the largest time attribute value in the second feature subset sequence, if the label is the second preset label, then skipping steps in the direction of decreasing time (using the second preset value as the step size); if the label is the first preset label, then gradually checking in the direction of increasing time until a turning point is found. The process of obtaining the second preset value is similar to the process of obtaining the first preset value, and will not be described again here.
[0065] In this embodiment, based on S310-S340, the target feature subset can be found quickly, thus improving the efficiency of obtaining the target feature subset.
[0066] In this embodiment, if there is no target feature subset of the user that satisfies the above conditions (i.e., the status label of the user corresponding to the feature subset whose time attribute value is less than the time attribute value of the target feature subset is the first preset label, and the status label of the user corresponding to the feature subset whose time attribute value is greater than the time attribute value of the target feature subset is the second preset label), then it is determined that positive samples cannot be extracted from the user's feature set, and the screening fails; if there is a target feature subset of the user that satisfies the above conditions, then it is determined that positive samples can be extracted from the user's feature set, and the screening succeeds.
[0067] S400, if the screening is successful, the target feature subset of the user is determined as the positive sample when training the target model; the target model is used to obtain the user's state label.
[0068] In one specific implementation, the method further includes: determining any subset of features whose time attribute value is less than the time attribute value of the target feature subset of the user as negative samples when training the target neural network model.
[0069] As a specific implementation method, the target model is the Extreme Gradient Boosting (XGB) model. Those skilled in the art will know that the structure and training process of the XGB model are existing technologies and will not be described in detail here.
[0070] In this embodiment, the selected target feature subset is used as a positive sample (representing state change points), and the feature subset preceding the target feature subset is used as a negative sample (representing unchanged state). This is used to train the target model, enabling the target model to learn the relationship between preset type feature data and the user's state, thus giving the target model the ability to predict the user's state label corresponding to any input feature subset. Those skilled in the art will understand that the process of training the target model is prior art and will not be described in detail here.
[0071] In this embodiment, for any user's feature set in the target user feature set, the feature subset with the median time attribute value is determined as the user's intermediate feature subset, the feature subset with the time attribute value less than the median value is determined as the user's first feature subset, and the feature subset with the time attribute value greater than the median value is determined as the user's second feature subset. If the user's status label corresponding to the intermediate feature subset is a first preset label, then the user's target feature subset is selected from the user's second feature subset; otherwise, the user's target feature subset is selected from the user's first feature subset. The target feature subset of each user is determined as the positive sample for training the target model. Compared with the prior art, this invention can not only automatically extract training samples, but also divide the user's feature set into a first feature subset and a second feature subset, and only select the target feature subset (positive sample) from the first feature subset or the second feature subset, which can reduce the number of features to be analyzed and improve the efficiency of training sample extraction.
[0072] Example 2:
[0073] This embodiment provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it performs the following steps:
[0074] Obtain the target user feature set; the target user feature set includes feature sets of several users, each feature set includes several feature subsets, and each feature subset includes several feature data of a preset type with the same time attribute.
[0075] For any user's feature set, the feature subsets whose time attribute values are equal to, less than, and greater than the median value are respectively determined as the user's intermediate feature subset, first feature subset, and second feature subset.
[0076] If the status label of a user corresponding to any intermediate feature subset is the first preset label, then the target feature subset of the user is selected from the second feature subset of the user; otherwise, the target feature subset of the user is selected from the first feature subset of the user. The target feature subset of the user satisfies the following: the status label of the user corresponding to the feature subset whose time attribute value is less than the time attribute value of the target feature subset is the first preset label, and the status label of the user corresponding to the feature subset whose time attribute value is greater than the time attribute value of the target feature subset is the second preset label.
[0077] If the screening is successful, the target feature subset of the user is determined as the positive sample when training the target model; the target model is used to obtain the user's state label.
[0078] Example 3:
[0079] This embodiment provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, it performs the following steps:
[0080] Obtain the target user feature set; the target user feature set includes feature sets of several users, each feature set includes several feature subsets, and each feature subset includes several feature data of a preset type with the same time attribute.
[0081] For any user's feature set, the feature subsets whose time attribute values are equal to, less than, and greater than the median value are respectively determined as the user's intermediate feature subset, first feature subset, and second feature subset.
[0082] If the status label of a user corresponding to any intermediate feature subset is the first preset label, then the target feature subset of the user is selected from the second feature subset of the user; otherwise, the target feature subset of the user is selected from the first feature subset of the user. The target feature subset of the user satisfies the following: the status label of the user corresponding to the feature subset whose time attribute value is less than the time attribute value of the target feature subset is the first preset label, and the status label of the user corresponding to the feature subset whose time attribute value is greater than the time attribute value of the target feature subset is the second preset label.
[0083] If the screening is successful, the target feature subset of the user is determined as the positive sample when training the target model; the target model is used to obtain the user's state label.
[0084] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.
[0085] While specific embodiments of the invention have been described in detail by way of example, those skilled in the art should understand that the examples are for illustrative purposes only and not intended to limit the scope of the invention. It should also be understood that various modifications can be made to the embodiments without departing from the scope and spirit of the invention. The scope of the invention is defined by the appended claims.
Claims
1. A method for processing user features, characterized in that, The method includes the following steps: Obtain the target user feature set; the target user feature set includes the feature sets of several users, each feature set includes several feature subsets, and each feature subset includes several feature data of the same time attribute of the preset type; the feature data of the preset type includes: age, gender, exercise frequency, blood sugar, blood pressure, blood lipids, drinking frequency, smoking frequency, medication feature data, imaging examination results, family history feature data, and past medical history feature data; For any user's feature set, the feature subsets whose time attribute values are equal to, less than, and greater than the median value are respectively determined as the user's median feature subset, first feature subset, and second feature subset; If the status label of a user corresponding to any intermediate feature subset is the first preset label, then the target feature subset of the user is selected from the second feature subset of the user; otherwise, the target feature subset of the user is selected from the first feature subset of the user; the target feature subset of the user satisfies the following: the status label of the user corresponding to the feature subset whose time attribute value is less than the time attribute value of the target feature subset is the first preset label, and the status label of the user corresponding to the feature subset whose time attribute value is greater than the time attribute value of the target feature subset is the second preset label; If the screening is successful, the target feature subset of the user is determined as the positive sample when training the target model; the target model is used to obtain the user's state label. The method further includes: identifying any subset of features whose time attribute value is less than the time attribute value of the target feature subset of the user as negative samples when training the target neural network model.
2. The user feature processing method according to claim 1, characterized in that, Filtering the user's target feature subset from the second feature subset includes: Sort the second feature subset in ascending order of time attribute values to obtain the first feature subset sequence; Using a first preset value as the step size, the judgment begins from the nth feature subset with the smallest time attribute value in the first feature subset sequence, where n is the first preset value; If the status label of the user corresponding to the smallest nth feature subset is the second preset label, then continue to determine the status label of the user corresponding to the (n-1)th feature subset with the smallest time attribute value in the feature subset sequence; If the status label of the user corresponding to the smallest (n-1)th feature subset is the first preset label, then the smallest (n-1)th feature subset is determined as the target feature subset; if the status label of the user corresponding to the smallest (n-1)th feature subset is the second preset label, then the judgment continues in the direction of decreasing time attribute value until the status label of the user corresponding to a certain feature subset is determined to be the first preset label, and the next feature subset of that feature subset is determined as the target feature subset.
3. The user feature processing method according to claim 2, characterized in that, The process of obtaining the first preset value includes: Obtain the difference between the maximum and minimum values of the time attributes of the feature subset included in the second feature subset; The average time interval of the time attribute values of adjacent feature subsets included in the second feature subset is obtained based on the difference and the number of feature subsets included in the second feature subset. The feature subset density per unit time period is obtained based on the average time interval. A first preset value is obtained based on the feature subset density; the first preset value and the feature subset density are positively correlated.
4. The user feature processing method according to claim 2, characterized in that, Filtering the user's target feature subset from the second feature subset also includes: If the status label of the user corresponding to the smallest nth feature subset is the first preset label, then continue to determine the status label of the user corresponding to the 2nth feature subset with the smallest time attribute value in the feature subset sequence; if the status label of the user corresponding to the smallest 2nth feature subset is the first preset label, then continue to determine in the direction of increasing time attribute value.
5. The user feature processing method according to claim 1, characterized in that, The process of obtaining the target user feature set includes: The initial user feature set is filtered according to preset filtering conditions to obtain the target user feature set. The initial user feature set includes the feature sets of several users. The preset filtering conditions include: the number of feature subsets included in the user's feature set is greater than or equal to the target preset number, and the difference between the maximum and minimum values of the time attributes of the feature subsets included in the user's feature set is greater than or equal to the target preset time interval.
6. The user feature processing method according to claim 1, characterized in that, The process of obtaining the target user feature set includes: For any feature subset of any user, if the feature data of a certain preset type in the feature subset of the user is empty, then the feature data of the preset type in the feature subset of the user is obtained based on the feature data of the same preset type in the similar feature subset of the user. The similar feature subset of the user satisfies the following condition: the time attribute value is less than the time attribute value of the feature subset, and the difference between the time attribute value and the time attribute value of the feature subset is less than or equal to a preset time threshold.
7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the user feature processing method as described in any one of claims 1 to 6.
8. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the user feature processing method as described in any one of claims 1 to 6.