Sequential access prediction method based on multi-task fusion access interval and related equipment

Abstract

The invention provides a sequential access prediction method based on a multi-task fusion access interval and related equipment, and is applied to the technical field of data processing. The method comprises the following steps: processing access information of a target user in a preset time period, and generating historical access time information of the target user, historical access record information of the target user and real-time access time information of the target user; processing the historical access time information of the target user, the historical access record information of the target user, the real-time access time information of the target user and the physiological feature information of the target user based on the target access result prediction model, and generating an interval prediction sub-task and a sequential access prediction sub-task; processing the interval prediction subtask and the sequential access prediction subtask to generate an initial prediction feature vector; and processing the initial prediction feature vector based on the access purpose information of the target user to generate target prediction information of the target user.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a sequential access prediction method based on multi-task fusion access interval and related equipment. Background Art

[0002] Sequential visit prediction is a prediction based on the user's historical visits, spanning various medical specialties and scenarios. With the advancement of technology, visit prediction research continues to develop and iterate. Early visit prediction models relied on heuristics and expert systems, such as the MYCIN system, which were limited by the high maintenance cost of manual rule curation. Advances in traditional machine learning methods have automated the process of training classifiers from electronic medical records, treating visit prediction as a multi-class classification task. Attention-based models, such as MMORE, MusaNet, and HiTANet, focus on capturing multi-scale temporal features. Although these models take into account the temporal information of user visits, they fail to effectively capture deeper information, such as the irregularity of visit intervals. How to solve the above problems is the focus of this solution.

[0003] It should be noted that the information disclosed in the above background technology section is only used to enhance the understanding of the background of the present disclosure, and therefore may include information that does not constitute the prior art known to ordinary technicians in the field. Summary of the invention

[0004] The purpose of this application is to provide a sequential access prediction method and related equipment based on multi-task fusion access intervals, which at least overcomes the problems existing in the prior art to a certain extent. Through the access interval prediction subtask and position encoding and interval encoding, it can fully capture the access interval and sequence information, make up for the shortcomings of previous research, more accurately reflect the fluctuations in user health status, and improve model performance. For example, ablation studies have shown that removing this task will reduce the accuracy. The multi-task learning paradigm is adopted to fuse the two subtasks, and the performance of each subtask is synergistically enhanced, so that the model can learn richer feature representations, effectively avoid overfitting, and improve the generalization ability in different tasks and data scenarios. The historical attention residual is introduced to improve the Transformer multi-head self-attention module to better capture long-term dependencies, which helps to extract key information from historical accesses and improve prediction accuracy.

[0005] Other features and advantages of the present application will become apparent from the following detailed description, or may be learned in part by the practice of the present invention.

[0006] According to one aspect of the present application, a sequential access prediction method based on multi-task fusion access interval is provided, including: obtaining access information of a target user within a preset time period, access purpose information of the target user, physiological characteristic information of the target user, an initial access result prediction model and a training sample set; processing the access information of the target user within the preset time period to generate historical access time information of the target user, historical access record information of the target user and real-time access time information of the target user; preprocessing the training sample set to generate a training sample set with target characteristic information; processing the initial access result prediction model based on the training sample set with target characteristic information to generate a target access result prediction model; processing the historical access time information of the target user, the historical access record information of the target user, the real-time access time information of the target user and the physiological characteristic information of the target user based on the target access result prediction model to generate an interval prediction subtask and a sequential access prediction subtask; processing the interval prediction subtask and the sequential access prediction subtask to generate an initial prediction feature vector; processing the initial prediction feature vector based on the access purpose information of the target user to generate target prediction information of the target user.

[0007] In another aspect of the present application, a sequential access prediction device based on multi-task fusion access interval includes: an acquisition module, which is used to acquire the access information of a target user within a preset time period, the access purpose information of the target user, the physiological characteristic information of the target user, an initial access result prediction model and a training sample set; a processing module, which is used to process the access information of the target user within the preset time period to generate the historical access time information of the target user, the historical access record information of the target user and the real-time access time information of the target user; pre-process the training sample set to generate a training sample set with target characteristic information; based on the training sample set with target characteristic information, the initial access result prediction model is processed to generate a target access result prediction model; based on the target access result prediction model, the historical access time information of the target user, the historical access record information of the target user, the real-time access time information of the target user and the physiological characteristic information of the target user are processed to generate an interval prediction subtask and a sequential access prediction subtask; the interval prediction subtask and the sequential access prediction subtask are processed to generate an initial prediction feature vector; based on the access purpose information of the target user, the initial prediction feature vector is processed to generate the target prediction information of the target user.

[0008] According to another aspect of the present application, an electronic device includes: a first processor; and a memory for storing executable instructions of the first processor; wherein the first processor is configured to implement the above-mentioned sequential access prediction method based on multi-task fusion access interval by executing the executable instructions.

[0009] According to another aspect of the present application, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a second processor, the method for sequential access prediction based on multi-task fusion access interval is implemented.

[0010] The present application provides a method and related equipment for predicting sequential access based on multi-task fusion access intervals. The server can fully capture the access interval and sequence information through the access interval prediction subtask and position encoding and interval encoding, make up for the shortcomings of previous research, more accurately reflect the fluctuations in user health status, and improve model performance. For example, ablation studies show that removing this task will reduce the accuracy. The multi-task learning paradigm is adopted to fuse the two subtasks, and the performance of each subtask is synergistically enhanced, so that the model can learn richer feature representations, effectively avoid overfitting, and improve the generalization ability in different tasks and data scenarios. The historical attention residual is introduced to improve the Transformer multi-head self-attention module to better capture long-term dependencies, which helps to extract key information from historical accesses and improve prediction accuracy.

[0011] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 A flowchart of a sequential access prediction method based on multi-task fusion access interval provided by an embodiment of the present application is shown;

[0013] Figure 2 A schematic diagram of the structure of a sequential access prediction device based on multi-task fusion access interval provided by an embodiment of the present application is shown;

[0014] Figure 3 A partial structural diagram of an access result prediction model provided by an embodiment of the present application is shown. DETAILED DESCRIPTION

[0015] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0016] Combine the following Figure 1To describe the sequential access prediction method based on multi-task fusion access interval according to an exemplary embodiment of the present application. It should be noted that the following application scenarios are only shown to facilitate understanding of the spirit and principles of the present application, and the embodiments of the present application are not limited in this regard. On the contrary, the embodiments of the present application can be applied to any applicable scenario.

[0017] In one implementation, the present application also proposes a sequential access prediction method based on multi-task fusion access intervals and related devices. Figure 1 The flowchart of a sequential access prediction method based on multi-task fusion access interval according to an embodiment of the present application is schematically shown. Figure 1 As shown, the method is applied to a server, comprising:

[0018] S101, obtaining access information of a target user within a preset time period, access purpose information of the target user, physiological feature information of the target user, an initial access result prediction model and a training sample set.

[0019] In one implementation, the target user is set to have performed 4 health-related activities in a medical institution in the past year (preset time period). The activity times are January 15, 2023, March 5, 2023, July 20, 2023, and October 10, 2023. Through these time records, the time intervals between adjacent activities can be calculated, such as the first to second activity interval of 49 days (March 5, 2023-January 15, 2023), the second to third activity interval of 137 days, and the third to fourth activity interval of 81 days. These time interval data will serve as an important basis for subsequent analysis of activity patterns and potential needs, similar to the treatment of adjacent medical time intervals in the article, for related task analysis. The first activity was in the comprehensive diagnosis and treatment area, and the content recorded was general physical discomfort, such as fatigue, mild headache, etc. The second activity was in the specialist diagnosis and treatment area, involving skin-related conditions, such as slight changes in local skin, itching, etc. The third activity returned to the comprehensive diagnosis and treatment area, showing some symptoms related to the digestive system, such as loss of appetite, occasional abdominal discomfort, etc. The fourth activity was in the sports rehabilitation area, mainly functional problems in specific parts of the body, such as limited waist movement, occasional numbness of the lower limbs, etc. The activity records and related condition descriptions in these different areas will be used as basic data for subsequent analysis and processing, similar to the construction of access list embedding representation, to provide clues for in-depth understanding of user health-related behaviors and trends.

[0020] The target user went to the comprehensive treatment area for the first time because he had not been feeling well overall recently and hoped that professionals would assess his physical condition and get some suggestions for improvement, which was for the purpose of health status assessment and consultation. The second time he went to the specialist treatment area was because of abnormal skin conditions, the purpose was to understand the causes of skin changes and get corresponding coping guidance. The third time he went to the comprehensive treatment area was because of some symptoms in the digestive system, hoping to get professional analysis and improvement suggestions, which was also for the purpose of health status assessment and consultation. The fourth time he went to the sports rehabilitation area was because of functional problems in specific parts of the body, which required professional assessment of the problem and seeking ways to improve the function. Assuming that after the third appearance of digestive system-related symptoms, professionals recommend that users have a reexamination after a period of time to observe changes in the overall physical condition, then when going to the sports rehabilitation area for the fourth time, part of the reason can also be regarded as a monitoring and follow-up of the previous changes in physical condition, that is, to check whether other parts were affected during the previous period of physical discomfort or whether the original discomfort caused new changes in physical function.

[0021] During each activity, the medical institution recorded some basic physiological data of the target user. For example, during the first activity, the body temperature was 37.2℃, the blood pressure was 118 / 78mmHg, and the heart rate was 78 times / minute; during the second activity, the body temperature was normal (36.7℃), the blood pressure was 116 / 76mmHg, and the heart rate was 76 times / minute; during the third activity, the body temperature was 36.5℃, the blood pressure was 128 / 82mmHg, and the heart rate was 77 times / minute; during the fourth activity, the body temperature was 36.6℃, the blood pressure was 122 / 78mmHg, and the heart rate was 80 times / minute. These physiological index data will be used as target physiological characteristic indicators to analyze the basic physical condition of the user and its changing trend, and combined with the access information to provide comprehensive data support for subsequent analysis and prediction. In addition to conventional physiological indicators, it also includes information such as height and weight. For example, the target user is 173cm tall and weighs 68kg during the first activity. During the second activity, due to skin conditions affecting living conditions, the weight dropped to 67kg. During the third activity, the weight returned to 68kg. During the fourth activity, due to functional problems in specific parts of the body affecting the amount of activity, the weight increased to 69kg. These data can also reflect changes in the user's health-related conditions from the side, and have a certain auxiliary effect on the overall analysis and prediction model.

[0022] like Figure 3As shown in the figure, in the initial access result prediction model, in the initial access result prediction model architecture, the access interval analysis subtask module is in a key position, focusing on in-depth analysis of the target user's health-related activity interval status. It constructs a time interval sequence by accurately calculating the time intervals of adjacent activities, which is the cornerstone of subsequent analysis. For example, if the user has health activity records on January 10, March 20, and June 5, 2023, the length of each adjacent interval is calculated in turn to form a time interval sequence such as [69 days, 77 days], which provides original data support for mining the user's health behavior rhythm. Coding strategy: Use position coding and interval coding to represent interval information. Position coding assigns a unique coding value based on the order of activities. The first activity is coded as 1, and the subsequent ones are incremented in sequence, so that the model can accurately grasp the logic of the activity order. Interval coding digitizes the interval duration according to preset rules, such as 0-30 days coded as 0.1, 31-60 days coded as 0.3, etc., to reflect the user's health stability or change trend. Short intervals may be due to urgent health issues that cause users to pay frequent attention, while long intervals imply relative stability or reduced attention. These coded values ​​are input into the underlying model operations as key features and are deeply integrated into the subsequent complex analysis and prediction process.

[0023] Health status trend prediction subtask module, information encoding and integration: This module focuses on predicting the trend of user health status. Comprehensively extract physical condition description information from the user's historical activity records, covering multiple contents such as fatigue, skin lesions, and abnormal digestive function, while integrating external professional health knowledge systems, such as knowledge related to common discomfort symptoms and potential serious disease risks. For example, for users who have mentioned digestive problems many times, with the help of the knowledge of the evolution path of digestive problems in the professional medical knowledge base, and the changes in physiological indicators of users' multiple examinations are analyzed in collaboration. The attention mechanism is used to intelligently screen and strengthen the weight of key information, and the massive information is condensed into a vector representation that can accurately reflect health trends, laying a solid data and knowledge foundation for predicting health trends. Trend prediction example: Taking a user who has been troubled by digestive problems for a long time as an example, the model gathers the symptom descriptions of his multiple medical records, the changes in gastroscopy and colonoscopy detection indicators, and the results of external digestive disease research. After complex calculations and logical inferences, it is predicted that his health status will gradually improve, maintain the status quo, or tend to deteriorate. This prediction provides key forward-looking guidance for user health management and precise allocation of medical resources, helping medical parties to plan intervention strategies in advance.

[0024] Private layer: Focusing on deep mining of individual features, the private layer plays a key role in the model architecture, focusing on in-depth analysis of the unique individual features of the target user. For each user, the private layer receives raw data inputs such as visit information, physiological feature information, and visit purpose information within a preset time period. It first performs deep cleaning and structural processing on the visit information. For example, for multiple health visit records of users, it accurately classifies and encodes them according to multiple dimensions such as time, visit department, and main symptoms. For physiological feature information, through complex statistical analysis and feature engineering technology, deep features such as physiological indicator fluctuation characteristics and multi-indicator correlation characteristics are extracted. The visit purpose information is converted into a semantically rich vector representation to accurately capture the motivation and expectations behind each user's visit. On this basis, the private layer uses advanced deep learning algorithms, such as deep belief networks (DBNs) or stacked autoencoders (SAEs), to perform nonlinear transformation and dimensionality reduction on these processed features, and mine low-dimensional potential feature representations hidden in high-dimensional data. These potential features are closely related to the user's individual health status, behavior patterns, and preferences, providing personalized basis for subsequent model accurate predictions. For example, for a user who suffers from a chronic disease and has adjusted his treatment plan multiple times due to fluctuations in his condition, the private layer can accurately capture the personalized pattern of his disease change cycle, drug response characteristics, and the impact of lifestyle adjustments on health, providing key decision-making basis for the model when predicting the user's future health trends, access needs, and the effectiveness of intervention measures, significantly improving prediction accuracy and personalized service levels, and making the model output closely fit the user's unique health trajectory and demand level.

[0025] Optimization of the sharing layer mechanism and expansion of feature fusion: The sharing layer is the core hub of model synergy and efficiency enhancement, and the feature fusion architecture is fully optimized. On the one hand, the feature vectors of the access interval and the health trend prediction subtask are closely connected to realize the interactive sharing of basic features; on the other hand, rich interactive items and innovative combination features are actively introduced. For example, the access interval feature is multiplied by the frequency feature of a specific symptom, and a new feature is constructed to reveal the frequency of symptoms under a specific interval, and the nonlinear relationship between health behavior and condition is deeply mined. If a symptom occurs frequently in a specific interval, the model can capture it keenly, injecting key insights into the accurate diagnosis of disease evolution and evaluation of intervention effects. Upgraded encoding structure: The multi-layer bidirectional transformer architecture is used to encode features, and the multi-head attention mechanism processes multi-feature associations in parallel. Each layer of the transformer contains multiple head attention units, each unit captures feature relationships from different perspectives, and comprehensively covers the information interaction dimension. At the same time, historical attention residuals are introduced to solve the problem of long-term dependence. When processing chronic disease user data, the model can trace back all relevant activity information in the past, break through the limitations of short-term data, accurately capture long-term health trajectories, and achieve long-term accurate prediction of health trends, providing excellent intelligent support for chronic disease management.

[0026] Decoding layer: used for accurate mapping and output of prediction information. As the key hub for the output of model prediction information, the decoding layer is responsible for accurately converting the intermediate results after complex processing and analysis into target prediction information with practical application value. After receiving the high-dimensional feature vector that integrates multi-task features and deep knowledge from the sharing layer, the decoding layer performs reverse deduction based on the predefined mapping rules and semantic conversion logic. For the interval prediction subtask, the decoding layer uses a decoding algorithm based on time series analysis, combined with the historical visit interval pattern and the current health status characteristics, to accurately predict the probability distribution of the time interval of the user's next health visit, and convert it into an easy-to-understand time interval prediction value, such as "the next visit time is expected to be within the next 45-60 days with a high probability". For the sequential visit prediction subtask, the potential visit sequence feature vector is decoded into a specific predicted access path and corresponding probability by combining the deep semantic understanding model with the medical field knowledge graph, such as "based on the current health trend, there is a 60% probability of going to a specialist clinic for in-depth examination, a 30% probability of re-examination and evaluation at a general clinic, and a 10% probability of participating in rehabilitation treatment consultation for the next visit."

[0027] When comprehensively generating target prediction information, the decoding layer fully considers the weight assigned to the user's visit purpose information and the confidence of different prediction results, and uses weighted fusion and decision logic optimization algorithms to output detailed, accurate and actionable prediction information such as "In view of the user's recent fluctuations in cardiovascular indicators and rehabilitation needs, it is predicted that the next visit will focus on cardiovascular specialist review (recommended examination items: cardiac ultrasound, dynamic electrocardiogram), and will also carry recent health monitoring data (daily records of blood pressure and heart rate). The purpose of the visit is to evaluate the efficacy of drugs and adjust the rehabilitation plan. The estimated visit duration is about 90 minutes." This provides strong decision-making support for user health management, early allocation of medical resources, and precise docking of medical services, and deeply empowers the intelligent upgrade of the entire process of medical and health services and the optimization of personalized experience.

[0028] Pre-training data collection adheres to the principles of extensiveness and accuracy. In addition to large-scale general health data sets, in-depth mining of health data subsets that match the target user's region, age, gender and other characteristics. For example, to predict the health trend of middle-aged female users in a specific region, the health data of the same age and gender in the region and the data of people with similar health risks are included, so that the model can deeply understand the health characteristics and laws of a specific user group, greatly improve the adaptability to the target user, and ensure that the prediction is close to the individual's actual situation. The data processing link strictly adheres to the quality checkpoint. First, the fuzzy health status description is strictly verified and labeled, such as "occasional chest discomfort" is refined into "mild chest tightness about 3 times a month, which lasts for about 5 minutes and can be relieved" to ensure data accuracy. Then, pre-processing is performed according to the health status category and activity type, distinguishing between sudden visits for acute diseases and regular review data for chronic diseases, helping the model to accurately learn various health patterns. When initializing parameters, the parameter values ​​are flexibly set according to the distribution of pre-training data and the initial characteristics of the target user. For users with a family history of cardiovascular disease, the initial parameters of cardiovascular risk indicators in similar cases are used as a reference to replace general initialization. The model accurately anchors individual characteristics at the start, accelerates convergence to the optimal solution, and ensures accurate and stable predictions.

[0029] To accurately drive model learning, a dedicated loss function is designed according to the characteristics of the subtask. The visit interval analysis subtask selects the mean square error loss function (MSELoss) to accurately measure the deviation between the predicted and actual intervals, and strongly constrains the model to accurately depict the temporal rhythm of health activities. The health status trend prediction subtask adopts the improved FocalLoss function, which flexibly allocates weights according to the importance of health trends and the difficulty of prediction, and increases the weight of deterioration trends, such as weighting the sensitive capture of early signs of weak deterioration in cancer, strengthening the model's ability to warn of key health risks, and guiding the model to focus on key trend predictions, avoiding interference from a large amount of normal state data, and improving the sensitivity of health risk monitoring. Hyperparameter optimization is the key to unleashing the potential of the model. Based on traditional learning rate, hidden layer dimension and other parameters, deep mining is carried out on hyperparameters related to visit intervals and health trends, such as visit interval encoding scaling factors, health trend attention weight parameters, etc. Cross-validation is used to evaluate model performance on multiple training subsets, such as cross-validation by age stratification, to find the optimal hyperparameter combination for each age group. After massive experimental iterations, we accurately lock in the hyperparameter set that adapts to a variety of scenarios, ensuring the optimal balance between model generalization and prediction accuracy, and robustly outputting reliable results in complex and changeable health prediction tasks, thus safeguarding the health of users.

[0030] S102, processing the access information of the target user within a preset time period to generate historical access time information of the target user, historical access record information of the target user and real-time access time information of the target user.

[0031] In one implementation, it is assumed that the target user has 5 relevant visit records on a health service platform in the past year (preset time period), and the following is the specific information: The first visit, the visit time was 10:30 on March 10, 2022, and the visit content was summarized as follows; the user felt that his physical condition was not good recently, specifically, he had discomfort in his head and his rest was not ideal, so he communicated with the consultant on the platform. The consultant asked about the duration of the relevant situation, the degree of daily feelings, and other issues, and gave some general life adjustment suggestions, such as regular work and rest, moderate relaxation, etc. The second visit, the visit time was 14:15 on May 5, 2022, and the visit content was summarized as follows: the user reported that the symptoms of head discomfort did not improve and there were some changes in vision. The specific circumstances of the visual changes were asked in detail, such as changes in visual clarity, whether there were abnormal light and shadow, etc., and the user was suggested to further evaluate the overall physical condition. The third visit, the visit time was 9:45 on July 20, 2022, and the visit content was summarized as follows: After a comprehensive physical examination (including scanning of specific parts, etc.), the user communicated with the consultant again. The fourth visit was at 11:20 on October 1, 2022. The visit content is summarized as follows: The user felt that his physical condition had further changed during the holiday, and urgently visited the platform. The consultant speculated that it was related to the internal pressure balance of the body based on previous records and the user's current description. The fifth visit was at 16:30 on December 15, 2022. The visit content is summarized as follows: The user reported that he had gone to a professional institution for examination and determined that a certain indicator inside the body was abnormal. This visit was to consult the consultant about the daily precautions and adjustment plan for this situation. The consultant planned a personalized lifestyle optimization plan for the user, including diet structure adjustment, daily activity planning, etc.

[0032] Calculate the time interval between adjacent visits:

[0033] The time interval between the first visit (10:30 on March 10, 2022) and the second visit (14:15 on May 5, 2022) is: First calculate the difference in days. March has 31 days and April has 30 days, so the interval is 56 days. Then calculate the time difference, 14:15-10:30 = 3 hours and 45 minutes.

[0034] The time interval from the second visit to the third visit (9:45 on July 20, 2022) is:

[0035] May has 31 days, June has 30 days, and the interval is 76 days. 9:45-14:15 = -4 hours and 30 minutes (the time difference here is negative, which means that the third visit time is earlier than the second visit time on the same day. When calculating the time interval, you can take the absolute value or process it according to specific needs). By analogy, calculate the time intervals of all adjacent visits. These interval information reflects the rhythm of users' attention and treatment of their health conditions at different stages, which can be used as part of the historical access time information. Access time series analysis: Arrange all access times in chronological order to form a time series: [10:30 on March 10, 2022, 14:15 on May 5, 2022, 9:45 on July 20, 2022, 11:20 on October 1, 2022, 16:30 on December 15, 2022]. Analyze the characteristics of this time series, for example, you can calculate the average number of days between visits and the distribution of access times (such as the proportion of visits on weekdays and weekends, etc.). Assuming that the average number of days between visits is about 60 days, and that users visit more frequently on weekdays, this suggests that users are more concerned about managing their health outside of work. These statistics and analysis results are all historical visit time information, which helps to understand users' health management behavior patterns.

[0036] Historical visit record information is generated, specifically: Detailed information of the first visit record: The user communicated with the health consultant due to head and rest problems. The consultant asked about the duration of physical discomfort (user feedback of about two weeks), the degree of daily feelings (described as intermittent head abnormality, moderate degree affecting daily life), the manifestation of rest problems (difficulty falling asleep, about 4-5 hours of sleep per night), etc. The consultant's advice is to adjust the work and rest time, avoid excessive fatigue, and moderately relax. Second visit record: The user's head discomfort has not been relieved and there are changes in vision. The visual health expert asked about the time of visual changes (started about a week ago), whether there are other abnormal visual feelings (no obvious eye pain but slight light and shadow perception), the association between head discomfort and visual changes (visual changes are more obvious when the head discomfort is aggravated), etc. The visual function examination items arranged include basic vision test, visual sensitivity test, etc. The preliminary examination results show that the vision fluctuates slightly. The expert recommends further evaluation of the overall physical condition to exclude potential related factors. In this way, the detailed records of each visit are sorted out, including the physical condition reported by the user, the inquiry content of the consultant or expert, the examination items and results (if any), and the adjustment suggestions given.

[0037] From the first visit to the last visit, the changes in the user's physical condition are sorted out: from the initial head discomfort and poor rest, to the head discomfort accompanied by visual changes, and then to the determination of abnormalities in certain indicators inside the body. At the same time, the changes in the consultant's or expert's response ideas and suggestions are sorted out: from the initial lifestyle adjustment, to the suspicion of physical function problems and the recommendation of a comprehensive examination, to further observation and analysis based on the examination results, and finally to the determination of abnormal conditions and the formulation of adjustment plans. For example, taking the symptom of head discomfort as an example, its severity varies in different visits (from intermittent abnormal sensations to aggravation), and related symptoms gradually appear (visual changes). The trajectories of the development of these physical conditions and the corresponding response processes constitute an important part of the historical visit record information, which is helpful for analyzing the evolution of the user's health status and the feedback on related suggestions.

[0038] Assume that at the current moment (15:00 on May 1, 2023), the user is accessing the health service platform, preparing to consult about the feedback and further optimization suggestions after the recent life adjustments under the previously determined abnormal internal body indicators. The real-time access time information at this time is 15:00 on May 1, 2023. This real-time access time is combined with the previous historical access time information to analyze whether the time pattern of user visits has changed, such as whether the platform is visited more frequently to consult health issues in a specific time period (such as after holidays, when the seasons change, etc.). At the same time, compared with the historical access record information, you can understand the relationship and development of the user's current visit with previous changes in physical condition and response suggestions. For example, this consultation is based on the needs of the continuous adjustment process after the abnormal internal body indicators were previously determined. Through real-time access time information, you can better grasp the dynamic process of user health management.

[0039] S103, preprocessing the training sample set to generate a training sample set with target feature information.

[0040] In one implementation, the training sample set is grouped to generate a grouped training sample set, wherein the grouped training sample set includes physiological characteristic information of users of different age groups and genders. Assume that we have a training sample set containing health-related data of 1,000 users, which records the basic information of the users (age, gender), various physiological indicators (such as blood pressure, heart rate, blood sugar, height, weight, etc.) and some living habit data (such as weekly exercise time, daily number of cigarettes smoked, drinking frequency, etc.). We divide age into three intervals: youth (18-35 years old), middle-aged (36-55 years old), and elderly (56 years old and above). Then further subdivide according to gender (male, female), so that six groups are obtained: young male group, young female group, middle-aged male group, middle-aged female group, elderly male group, and elderly female group. For example, user 1 (35 years old, male) belongs to the young male group; user 2 (42 years old, female) belongs to the middle-aged female group; user 5 (60 years old, male) belongs to the elderly male group.

[0041] The grouped training sample set is processed for feature extraction to generate an original feature library. For each group, we select some features related to the user's health status. In addition to the above-mentioned physiological indicators and living habits data, some derived features can also be calculated, such as body mass index (BMI = weight (kg) / height (m)2), blood pressure pulse pressure difference (systolic pressure - diastolic pressure), etc. Taking the young male group as an example, the BMI and pulse pressure difference of each user are calculated. For user 1 (height 175cm, weight 75kg), BMI = 75 / (1.752) ≈ 24.49, pulse pressure difference = 120-80 = 40. These original features (including original physiological indicators, living habits data and calculated derived features) are sorted out to form an original feature library. Each group has its own corresponding original feature library, which contains various feature information of the users in the group.

[0042] The original feature library is processed to generate training sets and validation sets. We decided to divide the original feature library of each group into training sets and validation sets at a ratio of 70%:30%. For example, if there are 200 users in the young male group, the data of 140 users will be used as the training set, and the data of 60 users will be used as the validation set. For the young male group, all feature data of 140 users are randomly selected as the training set, and the feature data of the remaining 60 users are used as the validation set. The same method is applied to the other five groups to generate grouped training sets and validation sets respectively. In this way, both the training set and the validation set retain the feature information of users of different age groups and genders, and the distribution is similar to the original sample set. The validation set is predicted based on the classifier to generate prediction results, and the training set is trained based on the preset algorithm to generate validation set class prediction results. The logistic regression classifier is selected as the example classifier (a more appropriate classifier can be selected according to the characteristics of the data in actual applications). The classifier is trained using the training set (including training data for each group). The goal of the training is to predict whether the user has potential health risks based on information such as the user's physiological characteristics and living habits (here it is simply defined as a binary classification problem, risky / no risk, and more complex classification goals can actually be defined based on specific research questions). For the validation set (which also includes validation data for each group), the feature data of each user is input into the trained classifier to obtain the prediction result. For example, for a user in the validation set of the young male group, the classifier predicts whether he has health risks based on his age, gender, blood pressure, heart rate, BMI, living habits and other characteristics. The prediction result is risky or no risk (assuming it is expressed in probability form, such as a risk probability of 0.6, indicating a higher risk).

[0043] At the same time, a preset algorithm (such as the K-nearest neighbor algorithm) is used to train the training set to obtain the validation set class prediction results. The K-nearest neighbor algorithm predicts the category of the validation set user (here also the category of health risks) based on the category of the neighbor users in the training set with similar characteristics to the users in the validation set. For example, for a user in the validation set, the K-nearest neighbor algorithm finds the K training set users with the most similar characteristics, and predicts the category of the validation set user based on the majority category of these K users, and obtains the validation set class prediction result (such as risky or non-risky).

[0044] The prediction results and the validation set prediction results are processed to generate target feature information, which is used to characterize the risk factors of abnormal physiological characteristics of users. The prediction results (such as risk probability, etc.) obtained based on the classifier are compared and analyzed with the prediction results of the validation set obtained based on the preset algorithm. For example, if the classifier predicts that a user has a higher risk (risk probability 0.7), and the K-nearest neighbor algorithm predicts that the user has no risk, then it is necessary to deeply analyze the reasons for this difference. It is found that some characteristics of the user (such as higher blood pressure, lower exercise time, and higher number of cigarettes) are given higher weights in the classifier, resulting in a higher risk probability, but in the K-nearest neighbor algorithm, due to the different feature distributions of its neighbor users, different prediction results are caused. These features related to the prediction difference (such as blood pressure, exercise time, number of cigarettes, etc.) are the risk factors that characterize the abnormal physiological characteristics of users. These factors are sorted out to form the target feature information. The target feature information can help us further understand which factors are most critical to the prediction of user health risks, thereby providing a basis for subsequent health management and intervention.

[0045] S104: Process the initial access result prediction model based on the training sample set with target feature information to generate a target access result prediction model.

[0046] In one implementation, it is assumed that we have completed the previous series of operations and obtained a training sample set with target feature information. The sample set contains rich information of many users. After processing and analysis, key features related to the physiological characteristics of users are extracted and associated with specific annotations or results, which are used to train the model to predict certain trends or results of users in health-related aspects (the specific content of the target feature information is not elaborated in detail here, only its existence form and function in the sample set are emphasized). The sample set contains user data of different age groups and genders to ensure that the model can learn diverse patterns. It is assumed that the initial access result prediction model adopts a multi-layer perceptron (MLP) structure, and its main components are as follows: The input layer is designed to receive various feature information related to the user, which comes from the training sample set with target feature information. The number of input nodes is determined according to the number of features contained in the training sample set, and each node corresponds to a feature, such as basic information of the user (age, gender, etc.) and other related features derived from the target feature information (the specific number and nature of features are determined by the previous data processing).

[0047] The model contains multiple hidden layers to achieve deep processing and feature learning of input data. The number of hidden layers and the number of neurons in each hidden layer can be adjusted according to actual needs and the complexity of the data. For example, three hidden layers are set, with 256 neurons in the first layer, 128 neurons in the second layer, and 64 neurons in the third layer. In the hidden layers, appropriate activation functions (such as ReLU functions) are used to introduce nonlinear factors so that the model can learn the complex nonlinear relationship between input features, so as to better fit the data and discover potential patterns. The structure of the output layer depends on the specific requirements of the prediction task. If the prediction is whether the user has a risk of abnormal physiological characteristics (assuming a binary classification problem), the output layer has only one neuron, and the Sigmoid activation function is used to map the output value to between 0 and 1, indicating the probability that the user belongs to a certain category (such as the probability of risk). If the prediction is more complex, such as multiple risk types or risk level grading, the number of neurons and activation function of the output layer will be adjusted accordingly.

[0048] First, the training sample set with target feature information is preprocessed. This includes data cleaning, removing invalid or seriously missing data records; data normalization, mapping the values ​​of different features to a specific interval (such as between 0 and 1 or between -1 and 1) to ensure that each feature has equal importance during the training process and avoid some features dominating the model training due to a large value range. Then, the processed sample set is randomly shuffled and divided into a training set and a validation set according to a certain ratio (such as 70% for training set and 30% for validation set). In order to improve the training efficiency, the training set is further divided into several small batches, each batch containing a certain number of samples (such as 64 samples per batch). Before the model starts training, the weights and bias parameters in the model need to be initialized. Random initialization methods are usually used, such as randomly extracting initial values ​​from a specific probability distribution (such as normal distribution or uniform distribution). The range of initial values ​​needs to be carefully selected to avoid problems such as gradient vanishing or gradient exploding in the early stages of training. For example, for weight parameters, they can be randomly initialized from a normal distribution with a mean of 0 and a standard deviation of 0.01, and the bias parameters can be initialized to 0.

[0049] In each training batch, the sample data is sequentially input into the input layer of the initial access result prediction model. The data enters the hidden layer through the input layer, where it undergoes a series of linear transformations (weights multiplied by input data) and nonlinear activation operations (through activation functions) to gradually extract and learn the feature representations in the data. Finally, after being processed by the hidden layer, the data reaches the output layer to obtain the model's predicted output for the sample. For example, for a sample containing the user's age, gender, and other target feature information, after the forward propagation calculation of the model, the output layer will give a predicted value, indicating the user's predicted result in a certain health-related aspect (such as risk probability, etc.). Based on the model's predicted output and the true label of the sample (from the training sample set), the loss value is calculated. The choice of loss function depends on the nature of the prediction task. If it is a classification task, the cross entropy loss function is used; if it is a regression task, the mean square error loss function is used, etc. Taking the binary classification task as an example, the binary cross entropy loss function is used to calculate the difference between the predicted value and the true label (0 or 1, indicating whether a certain situation exists). By calculating the loss function, we get a numerical value to quantify the inaccuracy of the model prediction. The larger the loss value, the greater the difference between the model's prediction result and the actual situation.

[0050] Based on the calculated loss value, the back propagation algorithm is used to calculate the contribution of each parameter (weight and bias) to the loss, that is, the gradient. The gradient descent algorithm is used to update the parameters of the model according to the direction and size of the gradient. During the update process, a pre-set learning rate is used to control the step size of the parameter update. The choice of learning rate is very critical. If the learning rate is too large, the model will not converge or even diverge; if the learning rate is too small, the training process will become very slow. For example, an adaptive learning rate strategy, such as Adagrad, Adadelta or Adam optimization algorithms, is used to dynamically adjust the learning rate according to the actual situation during the training process to achieve more effective parameter updates, so that the model gradually learns the rules in the data and reduces the prediction error. After each training cycle (epoch), the model is evaluated using the validation set. The evaluation indicators can be selected according to the specific requirements of the prediction task, such as accuracy, recall, F1 value (for classification tasks) or root mean square error (RMSE), mean absolute error (MAE) (for regression tasks), etc. By calculating these evaluation indicators on the validation set, the performance of the model on unseen data can be understood. If the performance indicators of the model on the validation set are no longer improving or are starting to decline, it means that the model is overfitting (the model has overlearned the training set data, resulting in a decrease in the generalization ability of new data). At this time, some measures can be taken to prevent overfitting, such as stopping training early, adding regularization terms (such as L1 regularization, L2 regularization), using Dropout technology, etc., to improve the generalization ability of the model.

[0051] Repeat the above process of forward propagation, loss calculation, back propagation and parameter update, and after multiple training cycles (dozens or even hundreds of epochs are required, the specific number depends on factors such as the complexity of the data, the structure of the model and performance requirements), continuously optimize the parameters of the model, so that the performance of the model on the training set and the validation set gradually improves until the preset stop condition is met (such as reaching the maximum number of training cycles, the performance of the validation set no longer improves, etc.). After the above complete training process, the initial access result prediction model continuously adjusts its internal weights and bias parameters, and gradually learns the complex relationship between the user characteristics and the prediction results contained in the training sample set with target feature information. When the model meets the stop condition, it becomes the target access result prediction model. This target model has the ability to accurately predict new user data, and can output corresponding prediction results (such as whether there is a risk of abnormal physiological characteristics and the degree of risk, etc.) based on the input user-related characteristics (including age, gender and other characteristics derived from the target feature information), and can generalize to a certain extent to user data that has not appeared in the training set, providing valuable prediction and analysis for practical application scenarios (such as health management, medical decision-making assistance, etc.).

[0052] More diversified evaluation indicators are used to comprehensively evaluate the model performance. For example, the mean absolute error (MAE) is used to evaluate the accuracy of visit interval prediction, and the health stability index (HSI) is used to measure the model's ability to predict the overall stability of the user's health status. HSI can be calculated by analyzing the degree of fluctuation of the user's health status over a period of time (combined with factors such as changes in physiological indicators and changes in physical condition descriptions). The lower the value, the more stable the health status. The model's ability to accurately predict HSI will help evaluate its effectiveness in user health management. During the verification process, a stratified sampling verification method is used to divide the data set into multiple levels based on factors such as the user's age, gender, and health status category, and then verify each level separately. This ensures that the effectiveness of the model in different types of user groups can be fully verified, avoiding verification bias caused by uneven data distribution. At the same time, the verification data set is regularly updated to include the latest user data to ensure that the model can always adapt to the changing health status and user behavior patterns, and maintain the accuracy and timeliness of its predictions.

[0053] S105, based on the target access result prediction model, the historical access time information of the target user, the historical access record information of the target user, the real-time access time information of the target user and the physiological feature information of the target user are processed to generate an interval prediction subtask and a sequential access prediction subtask.

[0054] In one embodiment, the historical access time information and the historical access record information of the target user are processed based on the target access result prediction model to generate historical access interval information and historical access consultation information, wherein the historical access interval information is used to characterize the access interval duration between different access times and the access duration of the current visit, and the historical access consultation information is used to characterize the access content and access direction of different access times. Based on the historical access interval information and the historical access consultation information, the real-time access time information and the physiological characteristic information of the target user are processed to generate the access prediction information of the target user, wherein the access prediction information of the target user is used to characterize the access content and access direction of the target user's current visit, and the access prediction information of the target user is processed to generate the interval prediction subtask and the sequential access prediction subtask.

[0055] Assume that the target user has 4 relevant visit records on a health management platform in the past period of time. The following are the details: First visit, visit time: January 5, 2023 10:00-10:30. Visit situation: The user has recently felt that his physical condition is not good, specifically manifested as easy fatigue and poor sleep, so he consulted the platform's health consultant. The consultant asked in detail about the user's daily life routine, including working hours, rest time arrangements, and daily activity. The user reported that his work is busy, he often needs to stay up late to work overtime, and he has fewer daily activities. Based on the user's description, the consultant recommends that the user adjust his work and rest schedule, try to ensure adequate sleep time, and appropriately increase some daily physical activities, such as walking, simple stretching exercises, etc., to improve the overall state of the body. The visit lasted 30 minutes.

[0056] The second visit was from 14:00 to 15:00 on March 10, 2023. Visit situation: The user said that the previous fatigue had not been relieved, and he began to feel slight discomfort in his head. This time, a general practitioner consultant on the platform was visited. The consultant asked detailed questions about the user's head discomfort symptoms, including the specific location of the pain, the frequency of the pain, and the approximate degree of the pain. At the same time, the consultant also conducted some basic physical examinations on the user, such as measuring the user's blood pressure and heart rate. The results showed that both blood pressure and heart rate were within the normal range. Based on the user's symptoms and examination results, the consultant recommends that the user continue to observe changes in the body. If the symptoms of head discomfort worsen further, a more in-depth physical examination should be conducted in time to determine whether there are any potential health problems. The visit lasted 60 minutes.

[0057] The third visit, visit time: 9:00-10:00 on June 1, 2023. Visit situation: The user reported that the discomfort in the head became more obvious, the pain level increased, and there were some changes in vision, and the vision felt a little blurred. This time, an ophthalmologist on the platform was visited. The expert first conducted a comprehensive examination of the user's eyes, including professional examination items such as vision test and intraocular pressure measurement. The examination results showed that the user's intraocular pressure was slightly higher than the normal range, and it was also learned that the user had been using electronic devices for a long time recently. Based on the examination results and the user's situation, the expert suspected that the blurred vision was caused by excessive use of the eyes, but did not rule out the possibility of other potential factors. The expert suggested that the user minimize the time of using electronic devices, use some eye drops to relieve eye fatigue, and arrange further brain-related examinations to rule out the possibility that brain problems affect vision. The visit lasted 60 minutes. The fourth visit, visit time: 11:00-12:30 on August 15, 2023. Visit: The user consulted the general practitioner again with the results of the previous brain examination, which showed some minor abnormalities in the brain blood vessels. The consultant, based on the user's previous series of physical discomfort symptoms and the results of this examination, recommended that the user undergo regular checkups and pay close attention to the overall condition of the body. At the same time, the consultant emphasized the importance of adjusting the lifestyle, including optimizing the diet structure, increasing nutritional intake, arranging exercise and exercise, and further adjusting the work and rest schedule to ensure adequate rest. The consultant also prescribed some nutritional supplements or auxiliary drugs that help improve brain blood circulation based on the user's specific situation. The visit lasted 90 minutes.

[0058] Calculate the interval between different visits: The interval from the first visit (January 5, 2023) to the second visit (March 10, 2023) is calculated as follows: the remaining days in January (31-5 = 26 days) + the number of days in February (2023 is not a leap year, February has 28 days) + 10 days in March = 64 days. The interval from the second visit to the third visit (June 1, 2023) is calculated as follows: the remaining days in March (31-10 = 21 days) + the number of days in April (30 days) + the number of days in May (31 days) + 1 day in June = 83 days. The interval from the third visit to the fourth visit (August 15, 2023) is calculated as follows: the remaining days in June (30-1 = 29 days) + the number of days in July (31 days) + 15 days in August = 75 days. Determine the duration of the current visit: According to the above description, the first visit duration is 30 minutes, the second visit duration is 60 minutes, the third visit duration is 60 minutes, and the fourth visit duration is 90 minutes. These visit intervals and the current visit duration together constitute the historical visit interval information, which is used for subsequent analysis of the rhythm and regularity of the user's health management behavior.

[0059] Historical visit consultation information: The first visit mainly revolves around the user's physical fatigue and sleep problems. The consultant asks about the user's living conditions and gives corresponding adjustment suggestions. The second visit focuses on the user's unrelieved fatigue and new symptoms of head discomfort. The consultant asks about the symptoms in detail and gives observation suggestions after a basic physical examination. The third visit is aimed at the user's aggravated head discomfort and new blurred vision problems. The ophthalmologist conducts a professional eye examination, asks about eye habits, and gives relief and further examination suggestions. The fourth visit is based on the user's brain examination results. The general practitioner consultant gives re-examination, lifestyle adjustment and auxiliary conditioning suggestions based on previous symptoms. The direction of the first visit is that the user seeks advice from the consultant to improve his living conditions due to physical discomfort, which belongs to the direction of basic health consultation. The direction of the second visit is to seek visits and suggestions from the general practitioner consultant for a deeper understanding of the physical condition for persistent physical discomfort, which is the direction of symptom follow-up. The direction of the third visit is to seek professional examinations and targeted solutions from ophthalmologists due to vision problems, focusing on the direction of solving eye health problems. The direction of the fourth visit is to seek comprehensive health management solutions from the general practitioner consultant based on the results of the brain examination, which is the direction of comprehensive health guidance. These visit contents and visit directions constitute historical visit consultation information, reflecting the development process and response methods of users' health problems.

[0060] Features are extracted from historical visit interval information, such as the average visit interval ((64+83+75) / 3≈74 days), the variance of the visit interval (calculation process omitted), and the average of the current visit duration ((30+60+60+90) / 4=60 minutes). Key information is extracted from historical visit consultation information, such as the frequency of physical discomfort symptoms (such as the occurrence of symptoms such as fatigue, head discomfort, blurred vision, etc. in different visits), the type of professional advice (the proportion of lifestyle adjustments, examination items, auxiliary conditioning, etc.). Assume that the current real-time visit time is 10:00 on November 1, 2023, and the target user has measured some physiological characteristics before this visit, such as blood pressure 130 / 85mmHg (slightly higher than the normal range) and heart rate 78 beats / minute (normal range). Input the historical visit data features, real-time visit time and physiological characteristics information into the target visit result prediction model. Based on the previously learned patterns and relationships, the model predicts that the content of this visit is related to previous changes in physical condition. Due to slightly high blood pressure and previous brain vascular conditions, it will involve further evaluation of the overall physical condition, especially the blood circulation system and body function, such as measuring blood pressure again and checking whether various physical indicators are stable. The direction of the visit is comprehensive health guidance, that is, professionals will comprehensively consider previous physical discomfort symptoms, examination results and current physiological characteristics, and provide a more comprehensive health management plan, including whether lifestyle auxiliary measures need to be adjusted and further examination of physical function items.

[0061] In the visit interval prediction subtask, position encoding and interval encoding are used to handle the irregularity of the visit interval. For the historical visit interval information of the target user, the visit time series is position encoded. For example, the first visit is encoded as position 1, the second visit is encoded as position 2, and so on. Then, for the visit interval duration, interval encoding is performed. Assume that the interval duration is mapped to a coding space according to certain rules, such as 0-30 days interval encoding is 0.1, 31-60 days interval encoding is 0.3, 61-90 days interval encoding is 0.5, and 90 days and above interval encoding is 0.8. For predicting the next visit interval, the model predicts the interval duration between the next visit and the current visit (November 1, 2023) based on the position encoding and interval encoding information of the historical visits, combined with the real-time physiological characteristics of the target user (such as blood pressure changes affect the visit interval, slightly higher blood pressure causes the user to pay more attention to health and shorten the visit interval) and other factors. For example, the model predicts that the next visit interval is between 40-60 days, the corresponding interval code is about 0.3, and the next visit is predicted to be in the 5th position (i.e., the order calculated from the first visit) based on the position code. The sequential visit prediction subtask is determined based on the historical visit consultation information and the visit prediction information. Based on the development sequence of physical discomfort symptoms in the historical visit consultation information (from fatigue to head discomfort to blurred vision to brain blood vessel-related conditions) and the changes in the visit direction (from basic health consultation to symptom follow-up, professional eye examination to comprehensive health guidance), as well as the content of this visit prediction (such as further assessment of overall physical condition), the model predicts the subsequent visit sequence and content.

[0062] For example, the model predicts that in the next visit, if the results of various physical indicators change further, more professional health experts (such as cardiovascular health experts or neurological health experts) will be visited for more in-depth evaluation and guidance, and the direction of the visit will be more focused on specific areas of physical function. If the examination results are relatively stable, regular review and lifestyle adjustment consultations will continue to be conducted at the general practitioner or health consultant. The direction of the visit is still comprehensive health guidance but the focus is on maintaining good physical condition. At the same time, the implementation of the recommendations of professionals in historical visits (such as whether lifestyle adjustment measures are taken on time, etc.) will also affect the prediction of the order and content of subsequent visits. If the user does not strictly implement the recommendations, it will lead to further changes in physical condition or new problems, thereby changing the order and content of the visit, such as early visits or increased visit frequency, and the content of the visit will also revolve around new problems.

[0063] In another embodiment, the real-time access time information of the target user and the physiological characteristic information of the target user are processed based on the historical access interval information and the historical access consultation information to generate access prediction information of the target user, including: processing the historical access interval information, the historical access consultation information, the real-time access time information of the target user and the physiological characteristic information of the target user to generate target characteristic information, wherein the target characteristic information includes the mean access interval duration, the variance of the access interval duration, the maximum access interval duration, the minimum access interval duration, the access keyword frequency and the target physiological characteristic index; processing the target characteristic information to generate a first characteristic association value and a second characteristic association value, wherein the first characteristic association value is used to characterize the correlation between the access interval duration pattern and the physiological characteristic change, and the second characteristic association value is used to characterize the correlation between the target access content and the real-time access time.

[0064] Assume that the target user has 5 health-related visits in the past year, and the visit times are: March 1, 2022, May 15, 2022, August 5, 2022, November 20, 2022, and February 10, 2023. Calculate the interval between adjacent visits: the interval between the first and second visits is 75 days (May 15-March 1), the interval between the second and third visits is 82 days (August 5-May 15), the interval between the third and fourth visits is 107 days (November 20-August 5), and the interval between the fourth and fifth visits is 82 days (February 10-November 20). Then the mean interval between visits is (75+82+107+82) / 4=86.4 days, and the variance calculation is (omitted). The maximum value is 107 days and the minimum value is 75 days.

[0065] The first visit mainly consulted about physical fatigue and poor sleep quality. The consultant asked about work intensity and daily routine, and suggested adjusting the routine and moderate exercise. The recorded keywords included "fatigue", "poor sleep", "adjustment of routine", "exercise", etc. The second visit was because fatigue did not ease and headaches occurred. The consultant asked about headache symptoms and family history, and suggested further observation. The keywords included "persistent fatigue", "headache", "family history", "observation", etc. The third visit was because the headache worsened and the vision was blurred. The ophthalmologist checked the vision and intraocular pressure, asked about eye habits, and suggested reducing eye use and using eye drops. The keywords included "worsened headache", "blurred vision", "eye habits", "eye drops", etc. The fourth visit brought the results of the brain examination (showing mild abnormalities in brain blood vessels). The consultant suggested regular reexamination, adjustment of diet and increase of exercise. The keywords included "abnormal brain blood vessels", "reexamination", "diet adjustment", "increased exercise", etc. The fifth visit was to consult about reexamination-related matters, and the keywords included "reexamination consultation". The frequency of each keyword was counted, such as "fatigue" appeared 2 times, "headache" appeared 3 times, etc.

[0066] Assume that the current real-time access time is 14:00 on May 1, 2023, and assume that the user measured blood pressure of 130 / 85mmHg (slightly higher than the normal range), heart rate of 78 beats / minute (normal range), and weight of 70kg (2kg more than the last visit) before the real-time access. These are used as target physiological characteristic indicators. The mean of the access interval has been calculated to be 86.5 days, the variance (assuming it is 120.25, the calculation process is omitted), the maximum value is 107 days, and the minimum value is 75 days. These characteristics reflect the time interval regularity of user visits. As shown in the above statistics, the word frequency of "fatigue" is 2, the word frequency of "headache" is 3, etc. These word frequency information can reflect the focus and changing trend of health issues that users are concerned about. Indicators such as blood pressure 130 / 85mmHg, heart rate 78 beats / minute, and weight 70kg reflect the user's current physiological state. Compared with the physiological characteristics of the previous visit (assuming that the previous blood pressure was in the normal range and the weight was 68kg), changes such as increased blood pressure and increased weight can be seen.

[0067] The method also includes a calculation formula for obtaining the first feature association value, the calculation formula is:

[0068]

[0069] Among them, i j represents the jth visit interval, which is a specific value to measure the time interval between two visits of a user, reflecting the time pattern of the user's visit; P j represents the physiological characteristic value at the jth visit, which represents the quantitative index of the user's physiological state at the time of the visit, such as heart rate value, blood sugar value, etc.; μ I Represents the mean of the access interval duration, which summarizes the average level of user access interval duration and helps provide a benchmark for subsequent calculation of covariance and feature association values; σ I Represents the standard deviation of the access interval, which is used to standardize the data when calculating the feature association value to make different features comparable; μ P and σ P Represent the mean and standard deviation of physiological characteristic values, respectively. The mean reflects the overall average level of physiological characteristic values, and the standard deviation measures the degree of dispersion of physiological characteristic value data, and participates in the calculation of covariance and the determination of characteristic association values. Cov(I,P) represents covariance, which indicates the linear correlation between the two variables of visit interval duration and physiological characteristic value. It is an intermediate variable for calculating characteristic association values. Its positive and negative sign and size reflect the consistency and correlation of the changing trends of the two variables. R IP Represents the first feature association value, which is between -1 and 1. The closer the absolute value is to 1, the stronger the correlation between the access interval duration pattern and the change of physiological characteristics. A positive value indicates a positive correlation, that is, the change trend of the access interval duration is consistent with that of the physiological characteristics. A negative value indicates a negative correlation, that is, the change trends of the two are opposite.

[0070] First feature association value (correlation between access interval duration pattern and physiological feature changes):

[0071] Assume that the visit interval time sequence is I = {75, 82, 107, 82}, and the corresponding physiological characteristic changes can be represented by the weight change sequence P = {0, 0, 2, 2} (assuming that there is no change in weight from the first to the second visit, and the weight increases by 2 kg from the third to the fifth visit). First, calculate the mean μ of I I =86.5, variance (calculated), the mean μ of P P =1, variance (Calculation process omitted). Calculate the covariance Cov(I,P) = 10.5 (calculation process omitted). Then the first characteristic correlation value (Hypothesis), this value is close to 1, indicating that the visit interval length pattern has a strong positive correlation with weight change, that is, weight tends to increase when the visit interval is longer.

[0072] The method further includes a calculation formula for obtaining a second feature association value, wherein the calculation formula is:

[0073]

[0074]

[0075] Among them, f kj represents the number of times keyword k appears in the jth access consultation information. It reflects the frequency of occurrence of a keyword in a specific access and is used to analyze the characteristics of the access content; t j represents the real time of the jth visit, which is the specific moment when the user visits, and is used to explore the relationship between the access content and the access time; μ Fk and σ Fk : Represent the mean and standard deviation of the number of keyword occurrences, respectively. The mean reflects the average level of the number of keyword occurrences in multiple visits, and the standard deviation describes the degree of dispersion of its occurrences, which is used to calculate the correlation; μ T and σ T Respectively represent the mean and standard deviation of real-time access time. The mean summarizes the average access time, and the standard deviation measures the dispersion of access time, and participates in the calculation of covariance and feature association value. Cov(F k ,T) represents covariance, which indicates the linear correlation degree between the two variables of keyword occurrence times and real-time access time. It is an intermediate step in calculating the feature correlation value, and its value reflects the synergy of the change trends of the two. R kTRepresents the second feature association value, which is also between -1 and 1. Its absolute value indicates the strength of the correlation between the target access content (keyword k) and the real-time access time. The positive and negative signs indicate the direction of the correlation. A positive correlation means that the number of keyword appearances and the access time have the same changing trend, while a negative correlation indicates the opposite changing trend.

[0076] Second feature association value (correlation between target access content and real-time access time): Taking the keyword "review" as an example, its occurrence sequence in the historical access consultation information is F = {0, 0, 0, 1, 1} (the fourth and fifth visits mentioned review), and the corresponding real-time access time sequence is T = {March 1, 2022, May 15, 2022, August 5, 2022, November 20, 2022, February 10, 2023}. Calculate the mean value μ of F F =0.4, variance (Calculation process omitted), the mean value μ of T T (The calculation is July 24, 2022, the specific calculation process is omitted). Calculate the covariance Cov(F k ,T)=100 (the specific calculation process is omitted). Then the second characteristic correlation value (Assuming σ T =100 days, the calculation result is). This value shows that the appearance of the keyword "review" has a certain correlation with the real-time access time, and the attention to review increases with time.

[0077] The first feature association value and the second feature association are processed to generate visit prediction information for the target user. According to the first feature association value, since the visit interval time pattern has a strong positive correlation with weight change, and the current weight has increased, combined with the real-time visit time, which is about 80 days from the last visit (close to the average interval), it is predicted that this visit will continue to focus on weight management and overall physical condition assessment, and will ask about more effective ways to control weight, such as adjusting the diet plan and increasing exercise intensity. From the second feature association value, "review" has a certain correlation with the visit time, and there have been abnormalities in brain blood vessels before. It is predicted that this visit will involve matters related to brain review, such as asking about review items and whether the review time interval needs to be adjusted. At the same time, due to the previous blurred vision problem, vision-related conditions will be mentioned again, such as whether the vision has changed and whether further examination is needed. Combined with the slightly high blood pressure in the physiological feature information, the forecast consultant will recommend further monitoring of blood pressure, such as regular blood pressure measurement, recording blood pressure changes, etc., and give corresponding lifestyle adjustment suggestions based on blood pressure conditions, such as reducing salt intake and increasing aerobic exercise.

[0078] The visit prediction information includes: in terms of visit content, the focus is on weight management, brain review, vision, blood pressure monitoring, etc.; in terms of visit direction, professionals will be sought for advice on comprehensive assessment of health status and corresponding adjustment plans, such as developing personalized diet and exercise plans, determining review plans, etc. For example, it is predicted that users will ask consultants questions such as "I have gained weight recently. How can I adjust my diet and exercise to control my weight? What items do I need to do in my brain review? My vision is still a little blurry. Do I need further examination? My blood pressure is slightly high. What should I pay attention to?" The consultant will give detailed answers based on the user's situation and provide specific suggestions and plans.

[0079] S106, processing the interval prediction subtask and the sequential access prediction subtask to generate an initial prediction feature vector.

[0080] In one implementation, it is assumed that the target user has 4 health-related visits in the past year, and the visit times are: April 1, 2022, June 15, 2022, September 20, 2022, and December 10, 2022. Calculate the interval between adjacent visits: the interval between the first and second visits is 75 days (June 15-April 1), the interval between the second and third visits is 97 days (September 20-June 15), and the interval between the third and fourth visits is 81 days (December 10-September 20). The mean of the visit interval is (75+97+81) / 3=84.33 days, and the variance calculation is (omitted). The maximum value is 97 days and the minimum value is 75 days. Perform position encoding on the access interval (assuming simple encoding, the first access position is 1, the second is 2, and the third is 3) and interval encoding (assuming 0-30 days are encoded as 0.1, 31-60 days are encoded as 0.3, 61-90 days are encoded as 0.5, and more than 90 days are encoded as 0.8, then 75 days are encoded as 0.5, 97 days are encoded as 0.8, and 81 days are encoded as 0.5). Construct a vector to represent the features of the interval prediction subtask, for example, V1 = [84.33, variance value (assuming 100), 97, 75, 0.5, 0.8, 0.5 (the variance value here is assumed to be 100, which actually needs to be accurately calculated, and more interval-related features can be added as needed).

[0081] Visit consultation information processing. The first visit mainly consulted about physical fatigue and sleep problems. The keywords were "fatigue" and "sleep problems". The consultant asked about work pressure, etc. and suggested adjusting the work and rest schedule, which can be represented as vector E O1 = [2,1,1] (assuming that the word frequency of "fatigue" is 2, the word frequency of "sleep problem" is 1, and the word frequency of "work pressure" is 1. The word frequency statistics here can adjust the weight according to the actual situation.) The second visit was because the fatigue was not relieved and there was a headache. The consultant asked about the headache symptoms and family history. The keywords were "persistent fatigue", "headache", and "family history", which are expressed as EO2 = [1,2,1]. At the third visit, the headache worsened and the vision was blurred. The ophthalmologist checked the vision and intraocular pressure. The keywords were "worsened headache", "blurred vision" and "intraocular pressure", which is represented by E O3 =[2,2,1]. The fourth visit brought the brain examination results (mild abnormalities), and the consultant recommended reexamination, adjustment of diet and exercise. The keywords are "brain abnormality", "reexamination", "diet adjustment" and "increased exercise", which is represented by E O4 =[1,2,2,2].

[0082] Calculate the rate of change of the embedding representation of adjacent accesses (here simply represented by vector difference), ΔE O2 =E O2 -E O1 =[-1, 1, 0], ΔE O3 =E O3 -E O2 =[1,0,0],ΔE O4 =E O4 -E O3 =[-1, 0, 1, 1]. Calculate the mean (assumed to be [0, 0.33, 0.33], the calculation process is omitted) and variance (assumed to be [0.67, 0.22, 0.22], the calculation process is omitted). Construct the feature vector of the sequential access prediction subtask, such as V O =[E O4 , mean vector (0, 0.33, 0.33) variance vector (0.67, 0.22, 0.22)] (the feature representation can be further optimized and expanded as needed, such as adding more features about access order and access change). O Fusion. A simple way is to directly concatenate to obtain the variance value of the initial prediction feature vector. Adjust the feature weights according to the trend of historical data. For example, if it is found that user health problems tend to be more serious when the access interval is longer in history (based on more data and analysis assumptions), and the current access interval is longer (97 days is close to the maximum value), the weights of interval-related features can be appropriately increased. For example, multiply the mean and maximum values ​​of the access interval in by a coefficient greater than 1 (assuming it is 1.2) to obtain the optimized initial prediction feature vector.

[0083] S107: Process the initial prediction feature vector based on the access purpose information of the target user to generate target prediction information of the target user.

[0084] In one implementation, the target user's visit purpose information is processed to generate visit type information. Assume that the target user has had the following health-related visits in the past period of time: The first visit was because the target user felt weak and tired recently, so he went to the health management center to consult about ways to improve his physical condition. The inquiries mainly focused on lifestyle adjustments and nutritional supplements. The health consultant gave suggestions on work and rest adjustments, diet improvements, and proper exercise. The second visit was because after adjusting the lifestyle for a period of time during the first visit, the fatigue was relieved but still existed, and mild dizziness symptoms occurred. This time, a general practitioner consultant was visited. The consultant conducted a basic physical examination (such as blood pressure, heart rate measurements, etc., and the results were basically normal), asked about detailed living habits and family medical history, and suspected that it might be an early symptom of potential chronic health. Further observation and regular check-ups are recommended. The third visit was because the dizziness symptoms persisted and there was occasional palpitations. The cardiovascular specialist consultant was visited. The consultant conducted a more in-depth cardiovascular system examination (such as electrocardiogram examination, etc., and found mild arrhythmia), and asked in detail about the amount of daily activities, work pressure, sleep quality, etc. The visit was mild cardiovascular dysfunction, and a drug intervention plan and further lifestyle adjustment suggestions were given, including reducing work pressure and increasing aerobic exercise. The fourth visit was after taking the medicine for a period of time. The palpitations were alleviated, but the patient still wanted to know how to better restore cardiovascular health. The rehabilitation expert was consulted. The expert developed a personalized rehabilitation exercise plan and diet nutrition matching plan based on the user's recovery, focusing on enhancing cardiovascular function and improving physical endurance.

[0085] Based on these visit purposes, we can classify them into different visit types: The first visit can be classified as "health consultation and lifestyle adjustment", the main purpose of which is to obtain general advice on improving physical condition. It does not involve health visits and interventions, and the focus is on prevention and health promotion. The second visit belongs to the "symptom follow-up and potential health screening" category. If the user's symptoms have not completely disappeared after adjusting their lifestyle and new symptoms have appeared, the consultant will conduct preliminary examinations and screenings to determine whether there are potential health risks. The third visit is the "health visit and specialist intervention" category. A specialist examination is conducted for clear symptoms (dizziness, palpitations), the visit results are obtained (mild cardiovascular dysfunction), and targeted intervention (drug intervention) is started. The fourth visit is the "rehabilitation guidance and health management category". During the intervention process, focus on how to better restore health and obtain professional guidance related to rehabilitation (rehabilitation exercise plan, diet combination).

[0086] The access type information is processed to generate the prediction weight coefficient. We set different access types to have different importance for prediction, and thus give different weight coefficients. For example, for the "health visit and specialist intervention" access, since it involves clear health visits and intervention plans, it has a greater impact on the subsequent health status, so a higher weight coefficient is given, assuming it is 0.4. The "rehabilitation guidance and health management" access is also critical for the user's current health recovery and long-term health management, and the weight coefficient is set to 0.3. The "symptom follow-up and potential health screening" access helps to discover potential problems in a timely manner, and the weight coefficient is set to 0.2. The "health consultation and lifestyle adjustment" access has a relatively small direct impact on the current health status, but it has a certain effect on overall health awareness and prevention, and the weight coefficient is 0.1. According to the above rules, for the target user's access type information, a prediction weight coefficient vector W = [0.1, 0.2, 0.4, 0.3] is generated, corresponding to the four types of access: "health consultation and lifestyle adjustment", "symptom follow-up and potential health screening", "health visit and specialist intervention", and "rehabilitation guidance and health management".

[0087] The initial prediction feature vector is processed based on the prediction weight coefficient to generate the target prediction information of the target user. Assume that we have obtained the initial prediction feature vector V = [v1, v2, ..., v n ](Here n is the dimension of the feature vector, and the specific value is determined according to the previous processing results of the interval prediction subtask and the sequential access prediction subtask. For example, n = 30. The elements in the vector include access interval duration features, keyword frequency features, access change features, etc. For example, v1 = 80 means that the average access interval duration is 80 days, and v5 = 0.6 means the frequency weighted value of a certain keyword, etc.) The target prediction information is calculated by weighted summation. Target prediction information vector P = [p1, p2, ..., p n ],in, here represents the i-th element in the initial prediction feature vector corresponding to the first visit type (“health consultation and lifestyle adjustment category”), Similarly, for other access types, ω1, ω2, ω3, and ω4 are prediction weight coefficients. For example, for i=1 (access interval mean feature), (Assuming that the average access interval duration corresponding to the second access type is The third time Fourth The calculation process is omitted).

[0088] The other elements in the vector P are calculated in the same way to obtain the target prediction information vector P. This vector can be used to further analyze the user's future health-related behaviors or conditions, such as predicting the possible time of the user's next visit, the types of health issues that the user may be concerned about, etc. If an element in is related to the visit time prediction and has a large value, it may mean that the user's next visit time is near; if an element is related to the weighted sum of specific health issue keywords and has a prominent value, it may prompt the user to focus on this health issue during the next visit.

[0089] In this application, the server can fully capture the access interval and sequence information through the access interval prediction subtask and position encoding and interval encoding, make up for the shortcomings of previous research, more accurately reflect the fluctuations in user health status, and improve model performance. For example, ablation studies show that removing this task will reduce the accuracy. The multi-task learning paradigm is adopted to fuse the two subtasks and synergistically enhance the performance of each subtask, so that the model can learn richer feature representations, effectively avoid overfitting, and improve the generalization ability in different tasks and data scenarios. The historical attention residual is introduced to improve the Transformer multi-head self-attention module to better capture long-term dependencies, which helps to extract key information from historical visits and improve prediction accuracy.

[0090] In one embodiment, if Figure 2 As shown, the present application also provides a sequential access prediction device based on multi-task fusion access interval, comprising:

[0091] An acquisition module 201 is used to acquire the target user's access information within a preset time period, the target user's access purpose information, the target user's physiological feature information, the initial access result prediction model and the training sample set;

[0092] The processing module 202 is used to process the access information of the target user within a preset time period to generate the historical access time information, historical access record information and real-time access time information of the target user; pre-process the training sample set to generate a training sample set with target feature information; process the initial access result prediction model based on the training sample set with target feature information to generate a target access result prediction model; process the historical access time information, historical access record information, real-time access time information and physiological feature information of the target user based on the target access result prediction model to generate an interval prediction subtask and a sequential access prediction subtask; process the interval prediction subtask and the sequential access prediction subtask to generate an initial prediction feature vector; process the initial prediction feature vector based on the access purpose information of the target user to generate target prediction information of the target user.

[0093] Each embodiment in the present application is described in a related manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for evaluating the sequential access prediction method based on multi-task fusion access interval, electronic device, electronic device, and readable storage medium embodiment, since it is basically similar to the above-mentioned embodiment of the sequential access prediction method based on multi-task fusion access interval, the description is relatively simple, and the relevant parts can be referred to the partial description of the embodiment of the sequential access prediction method based on multi-task fusion access interval.

Claims

1. A sequential access prediction method based on multi-task fusion access interval, characterized in that: include: Obtaining the target user's access information within a preset time period, the target user's access purpose information, the target user's physiological feature information, the initial access result prediction model and the training sample set; Processing the access information of the target user within a preset time period to generate historical access time information of the target user, historical access record information of the target user and real-time access time information of the target user; Preprocessing the training sample set to generate a training sample set with target feature information; Processing the initial access result prediction model based on the training sample set with target feature information to generate a target access result prediction model; Based on the target access result prediction model, the historical access time information of the target user, the historical access record information of the target user, the real-time access time information of the target user and the physiological feature information of the target user are processed to generate an interval prediction subtask and a sequential access prediction subtask; Processing the interval prediction subtask and the sequential access prediction subtask to generate an initial prediction feature vector; The initial prediction feature vector is processed based on the access purpose information of the target user to generate target prediction information of the target user.

2. The method according to claim 1, characterized in that Preprocessing the training sample set to generate a training sample set with target feature information includes: Performing grouping processing on the training sample set to generate a grouped training sample set, wherein the grouped training sample set includes physiological characteristic information of users of different age groups and genders; Perform feature extraction on the grouped training sample set to generate an original feature library; Processing the original feature library to generate a training set and a verification set; Performing prediction processing on the verification set based on the classifier to generate a prediction result; Performing training processing on the training set based on a preset algorithm to generate a validation set class prediction result; The prediction result and the validation set prediction result are processed to generate target feature information, where the target feature information is used to characterize risk factors for abnormal physiological characteristics of the user.

3. The method according to claim 1, characterized in that Based on the target access result prediction model, the historical access time information of the target user, the historical access record information of the target user, the real-time access time information of the target user, and the physiological feature information of the target user are processed to generate an interval prediction subtask and a sequential access prediction subtask, including: Based on the target access result prediction model, the historical access time information and the historical access record information of the target user are processed to generate historical access interval information and historical access consultation information, wherein the historical access interval information is used to characterize the access interval duration between different access times and the access duration of the current time, and the historical access consultation information is used to characterize the access content and access direction of different access times; Based on the historical access interval information and the historical access consultation information, the real-time access time information of the target user and the physiological characteristic information of the target user are processed to generate access prediction information of the target user, wherein the access prediction information of the target user is used to characterize the access content and access direction of the target user's current visit; The access prediction information of the target user is processed to generate an interval prediction subtask and a sequential access prediction subtask.

4. The method according to claim 3, characterized in that The real-time access time information of the target user and the physiological characteristic information of the target user are processed based on the historical access interval information and the historical access consultation information to generate access prediction information of the target user, including: Processing the historical access interval information, the historical access consultation information, the real-time access time information of the target user and the physiological characteristic information of the target user to generate target characteristic information, wherein the target characteristic information includes the mean value of the access interval duration, the variance of the access interval duration, the maximum value of the access interval duration, the minimum value of the access interval duration, the frequency of access keywords and the target physiological characteristic index; Processing the target feature information to generate a first feature association value and a second feature association value, wherein the first feature association value is used to characterize the correlation between the access interval duration pattern and the physiological feature change, and the second feature association value is used to characterize the correlation between the target access content and the real-time access time; The first feature association value and the second feature association are processed to generate access prediction information of the target user.

5. The method according to claim 4, characterized in that The real-time access time information of the target user and the physiological characteristic information of the target user are processed based on the historical access interval information and the historical access consultation information to generate access prediction information of the target user, further comprising: The method further includes a calculation formula for obtaining a first feature association value, wherein the calculation formula is: Among them, i j represents the jth visit interval; P j represents the physiological characteristic value at the jth visit; μ I represents the mean of the access interval; σ I represents the standard deviation of the visit interval; μ P and σ P Represent the mean and standard deviation of physiological characteristic values ​​respectively; Cov(I,P) represents covariance; R IP represents the first feature association value; The method further includes a calculation formula for obtaining a second feature association value, wherein the calculation formula is: Among them, f kj represents the number of occurrences of keyword k in the jth access consultation information; t j represents the real time of the jth visit; and Respectively represent the mean and standard deviation of the number of keyword occurrences; μ T and σ T Represent the mean and standard deviation of real-time access time respectively; Cov(F k ,T) represents covariance; R kT Represents the second feature association value.

6. The method according to claim 3, characterized in that Processing the interval prediction subtask and the sequential access prediction subtask to generate an initial prediction feature vector includes: Performing feature extraction processing on the interval prediction subtask to generate access time interval sequence features, access interval number features and access interval information features; Processing the access interval information features to generate adjacent interval difference sequence information, difference sequence mean, and difference sequence variance; Performing dimensionality reduction processing on the sequential access prediction subtask to generate a one-dimensional prediction feature vector; Performing feature extraction processing on the one-dimensional prediction feature vector to generate target access list information, a mean of a similarity sequence of adjacent access information, and a variance of a similarity sequence of adjacent access information; The access time interval sequence feature, the access interval number feature, the target access list information, the mean of the adjacent access information similarity sequence and the variance of the adjacent access information similarity sequence are processed to generate an initial prediction feature vector.

7. The method according to claim 6, characterized in that The initial prediction feature vector is processed based on the access purpose information of the target user to generate target prediction information of the target user, including: Processing the target user's access purpose information to generate access type information; Processing the access type information to generate a prediction weighting coefficient; The initial prediction feature vector is processed based on the prediction weight coefficient to generate target prediction information of the target user.

8. A sequential access prediction device based on multi-task fusion access interval, characterized in that: The device comprises: An acquisition module, used to acquire the target user's access information within a preset time period, the target user's access purpose information, the target user's physiological feature information, the initial access result prediction model and the training sample set; A processing module is used to process the access information of the target user within a preset time period to generate the target user's historical access time information, the target user's historical access record information and the target user's real-time access time information; pre-process the training sample set to generate a training sample set with target feature information; based on the training sample set with target feature information, process the initial access result prediction model to generate a target access result prediction model; based on the target access result prediction model, process the target user's historical access time information, the target user's historical access record information, the target user's real-time access time information and the target user's physiological feature information to generate an interval prediction subtask and a sequential access prediction subtask; process the interval prediction subtask and the sequential access prediction subtask to generate an initial prediction feature vector; process the initial prediction feature vector based on the target user's access purpose information to generate the target prediction information of the target user.

9. An electronic device, characterized in that: include: a first processor; and a memory for storing executable instructions of the first processor; The first processor is configured to execute the sequential access prediction method based on multi-task fusion access interval described in any one of claims 1 to 7 by executing the executable instructions.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the second processor, the sequential access prediction method based on multi-task fusion access interval described in any one of claims 1 to 7 is implemented.

Images