Nursing equipment recommendation method and system based on user historical behaviors

Abstract

The invention discloses a nursing equipment recommendation method and system based on user historical behaviors, and relates to the technical field of equipment recommendation. Behavior patterns of users are identified, the users are divided into different groups based on a clustering result, user behavior data similarity is described by using a semi-variance function to form a semi-variance model; after a corresponding vacancy filling model is selected based on semi-variance model features, the selected vacancy filling model is trained according to the clustered user behavior data so as to estimate behavior data missing in time or space, the weight of each feature data is distributed in combination with the feature data of the user behavior data and a prediction result of the vacancy filling model, and the weight of each feature data is calculated. And setting a recommendation rule by using a multi-factor weighting method, and generating a nursing equipment recommendation list based on the recommendation rule. The recommendation system can utilize known data points to estimate missing values, so that the problem of data discontinuity or missing is solved, the space-time gap of user behavior data is filled after user behaviors are predicted, and the recommendation effect and accuracy are effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of equipment recommendation, and in particular to a method and system for recommending nursing equipment based on user historical behavior. Background Art

[0002] With the arrival of an aging society, the demand for nursing equipment has increased significantly. The demand for smart nursing equipment by the elderly and patients with chronic diseases is not limited to basic health monitoring, but also includes personalized support, such as assisted walking devices, smart mattresses, health management equipment, etc. In order to improve the adaptability of equipment, recommendation methods based on user behavior are becoming increasingly important.

[0003] The prior art has the following defects:

[0004] 1. Existing recommendation systems rely on complete user behavior data to ensure the accuracy of device recommendations. However, due to the limitations of behavior data collection, data discontinuity or missing time periods may occur, resulting in reduced recommendation effectiveness.

[0005] 2. Recommendation systems often find it difficult to accurately capture users' personalized needs, especially when processing behavioral data with temporal and spatial changes. Insufficient data sampling may prevent them from identifying the dynamic changes in user needs, resulting in inaccurate recommendations.

[0006] Based on this, the present invention proposes a nursing equipment recommendation method and system based on user historical behavior, which can use known data points to estimate missing values, thereby making up for the problem of discontinuous or missing data, and after predicting user behavior, fill the temporal and spatial gaps in user behavior data, effectively improving the recommendation effect and accuracy. Summary of the invention

[0007] The purpose of the present invention is to provide a nursing equipment recommendation method and system based on user historical behavior to address the deficiencies in the background technology.

[0008] In order to achieve the above object, the present invention provides the following technical solution: a nursing device recommendation method based on user historical behavior, the recommendation method comprising the following steps:

[0009] Collect user behavior data, add time and space labels to the behavior data, use the K-means algorithm to perform cluster analysis on the user behavior data, identify user behavior patterns, and divide users into different groups based on the clustering results;

[0010] The semivariance function is used to describe the similarity of user behavior data to form a semivariance model. After selecting the corresponding gap filling model based on the semivariance model features, the selected gap filling model is trained based on the clustered user behavior data to estimate the behavior data missing in time or space.

[0011] Combine the characteristic data of user behavior data and the prediction results of the gap-filling model, assign weights to each characteristic data, use the multi-factor weighted method to set recommendation rules, and generate a nursing equipment recommendation list based on the recommendation rules.

[0012] In a preferred embodiment, using a semivariance function to describe the similarity of user behavior data to form a semivariance model includes the following steps:

[0013] In the user behavior data, select the feature pairs that need to describe the similarity. For each pair of user data points, calculate the difference between the feature values. The expression is: ΔX = X i -X j , where ΔX is the difference, X i is the data point of user i, X j is the data point of user j, and the square of the difference is used to get the variation value, which is expressed as: In the formula, r is the variation value;

[0014] The user location data is divided into different distance intervals, and the user data pairs are grouped according to the distance interval. Each group contains all the user pairs in the distance interval. The mean square variation value of all data pairs in each distance interval is calculated as the semivariance value of the distance to complete the construction of the semivariance model. The semivariance model expression is: Where R is the semivariance value, N(h) represents the number of data pairs in the distance interval, and the semivariance values ​​at different distances are plotted to obtain the semivariogram, which is used to describe the relationship between distance and variation.

[0015] In a preferred embodiment, after selecting the corresponding gap filling model based on the semivariance model feature, the selected gap filling model is trained according to the clustered user behavior data to estimate the behavior data missing in time or space, including the following steps:

[0016] Based on the clustering results, the cluster data is extracted as the training set, and the gap filling model is trained on the selected cluster data;

[0017] The location of the behavior data missing in time or space is estimated by the gap filling model, and the gap filling model expression is: In the formula, is the predicted value at position s0, that is, the missing behavior data value to be estimated, X(s i ) is the position s i The observed value of the known data point at position s i The behavior data value at, N represents the total number of known data points used for interpolation, λ i Indicates position s i For the interpolation weight of position s0, the interpolation result is filled into the blank area of ​​the user behavior data to generate a complete historical record.

[0018] In a preferred embodiment, the interpolation weight λ i The acquisition logic is: Get location s i The observed value fluctuation amplitude and position s i The semivariance value and semivariance sill value between the position s0 are normalized to map the value range of the observation value fluctuation amplitude, semivariance value and semivariance sill value to [0,1]. The observation value fluctuation amplitude, semivariance value and semivariance sill value after normalization are summed to obtain the position s i Sum the value, take the inverse of the sum as the influence value, sum the influence values ​​of all positions to get the influence denominator value, and divide the influence value by the influence denominator value to get the influence value of each position s i The interpolation weight λ i .

[0019] In a preferred embodiment, cluster analysis is performed on user behavior data using a K-means algorithm to identify user behavior patterns, including the following steps:

[0020] Randomly select K data points as the initial cluster centers, calculate the distance between each data point and the cluster center, and assign the data point to the cluster with the smallest distance;

[0021] Recalculate the cluster center for each cluster and set the cluster center to the mean of all data points in the cluster. When the cluster center no longer changes or reaches the preset number of iterations, the algorithm converges.

[0022] According to the clustering results, analyze the characteristic values ​​of each cluster center, extract typical behavior patterns, and name each cluster group;

[0023] According to the clustering results, each user is assigned to a similar group, and the number of users and feature distribution of each group are calculated to obtain the proportion and feature distribution of different groups in the overall users.

[0024] In a preferred embodiment, the characteristic data of the user behavior data and the prediction results of the gap filling model are combined to assign weights to each characteristic data, and a multi-factor weighted method is used to set recommendation rules. Based on the recommendation rules, a nursing device recommendation list is generated, including the following steps:

[0025] Set recommendation rules based on weighted feature data and the results of the gap-filling model;

[0026] According to the recommendation rules, a recommendation list is generated based on the feature score of each device, and the comprehensive score of each nursing device is calculated: In the formula, S i is the comprehensive score of the i-th nursing device, m is the number of feature data, and wj is the weight of the jth feature data, F ij For the performance of the i-th nursing device on the j-th feature data, all nursing devices are sorted from large to small according to the comprehensive score to generate a list of recommended devices;

[0027] The recommended nursing devices are displayed to users according to priority, including the characteristics of each device, applicable population and usage effect information.

[0028] In a preferred embodiment, the recommendation system collects user behavior data and adds time and space tags to the behavior data, including the following steps:

[0029] Through smart wearable devices and smart monitoring devices, users' health data, including blood pressure, heart rate, body temperature, sleep quality, and activity level, are collected in real time, and the frequency and duration of users' use of various care devices are recorded;

[0030] Add a label of the user's time period to each piece of behavior data to analyze the user's health changes and device usage habits in different time periods, and mark the date of the behavior data to capture the periodic changes in the behavior data;

[0031] Geographic location information is added to behavioral data to analyze the impact of different locations on health data and device usage, and activity area labels are added to users to analyze health status and device requirements in different activity areas.

[0032] Nursing equipment recommendation system based on user historical behavior, including data collection module, clustering module, model training module, and recommendation module;

[0033] Data collection module: collects user behavior data and adds time and space tags to the behavior data;

[0034] Clustering module: Use K-means algorithm to perform cluster analysis on user behavior data, identify user behavior patterns, and divide users into different groups based on clustering results;

[0035] Model training module: Use the semivariance function to describe the similarity of user behavior data to form a semivariance model. After selecting the corresponding gap filling model based on the semivariance model features, the selected gap filling model is trained based on the clustered user behavior data to estimate the behavior data missing in time or space.

[0036] Recommendation module: Combine the characteristic data of user behavior data and the prediction results of the gap-filling model, assign weights to each characteristic data, use the multi-factor weighted method to set recommendation rules, and generate a recommended list of nursing equipment based on the recommendation rules.

[0037] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0038] The present invention uses K-means algorithm to cluster user behavior data, identify user behavior patterns, divide users into different groups based on clustering results, use semivariance function to describe user behavior data similarity to form a semivariance model, select the corresponding gap filling model based on the semivariance model feature, train the selected gap filling model based on clustered user behavior data to estimate behavior data missing in time or space, combine the feature data of user behavior data and the prediction results of the gap filling model, assign weights to each feature data, use multi-factor weighting method to set recommendation rules, and generate a nursing equipment recommendation list based on the recommendation rules. The recommendation system can use known data points to estimate missing values, thereby making up for the problem of discontinuous or missing data, and after predicting user behavior, fill the time and space gaps of user behavior data, effectively improving the recommendation effect and accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0040] Figure 1 The figure is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0041] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0042] Example 1: Please refer to Figure 1 As shown, the nursing device recommendation method based on user historical behavior described in this embodiment includes the following steps:

[0043] The recommendation system collects user behavior data, including health parameters (such as blood pressure, heart rate, body temperature, etc.) and the frequency and duration of using nursing equipment. Time and space labels (such as time of day, geographic location, etc.) are added to the behavior data. The K-means algorithm is used to perform cluster analysis on user behavior data to identify users' typical behavior patterns, such as the frequency of getting up at night and the amount of exercise. Based on the clustering results, users are divided into different groups so that the behavioral characteristics of similar people can be considered when making recommendations, thereby improving the accuracy and applicability of recommendations. The semivariance function is used to describe the similarity of user behavior data to form a semivariance model. Based on the semivariance function, the semivariance model is used to describe the similarity of user behavior data. After the variance model features are selected to select the corresponding gap-filling model, the gap-filling model includes a common gap-filling model, a general gap-filling model, etc. The selected gap-filling model is trained based on the clustered user behavior data to estimate the behavior data that is missing in time or space. The interpolation result can fill in the blank areas of the user behavior history records and generate more complete behavior data. The weight of each feature data is assigned by combining the feature data of the user behavior data (such as sleep quality, daytime activity, frequency of device use, etc.) and the prediction results of the gap-filling model. The recommendation rules are set using the multi-factor weighted method. A recommendation list of nursing equipment is generated based on the recommendation rules, and the list is displayed in order of priority.

[0044] This application uses the K-means algorithm to perform cluster analysis on user behavior data, identify user behavior patterns, divide users into different groups based on clustering results, use semivariance functions to describe the similarity of user behavior data to form a semivariance model, select the corresponding gap filling model based on the semivariance model features, train the selected gap filling model based on the clustered user behavior data to estimate the behavior data missing in time or space, combine the feature data of the user behavior data and the prediction results of the gap filling model, assign weights to each feature data, use a multi-factor weighted method to set recommendation rules, and generate a nursing equipment recommendation list based on the recommendation rules. The recommendation system can use known data points to estimate missing values, thereby making up for the problem of discontinuous or missing data, and after predicting user behavior, fill the time and space gaps in user behavior data, effectively improving the recommendation effect and accuracy.

[0045] Example 2: The recommendation system collects user behavior data, which includes health parameters (such as blood pressure, heart rate, body temperature, etc.) and the frequency and duration of using nursing equipment, and adds time and space tags (such as time of day, geographic location, etc.) to the behavior data, including the following steps:

[0046] Health parameter monitoring: Collect users’ health data in real time through smart wearable devices, smart monitoring devices, etc., including blood pressure, heart rate, body temperature, sleep quality, activity level, etc. These data can be collected regularly (such as every hour or every day) and transmitted to the recommendation system in real time.

[0047] Equipment usage frequency and duration records: Record the frequency (such as the number of times per day) and duration (such as the duration of each use) of various types of care equipment used by users to form detailed usage data.

[0048] Daily time period labels: Add labels of the user's time period (such as morning, afternoon, evening, etc.) to each behavioral data to facilitate analysis of the user's health changes and device usage habits in different time periods.

[0049] Daily, weekly, and monthly periodicity: Annotate the date and day of the week of the behavior data to capture possible periodic changes. For example, weekends may have different behavior patterns, and long-term monitoring can also capture monthly or quarterly changes.

[0050] Special event tags: Add special event tags to behavioral data during important nodes (such as user birthdays, holidays) or special conditions (such as recurrence of a disease) to facilitate the system to identify and handle behavioral anomalies in these scenarios.

[0051] Geolocation tagging: Adding geolocation information to behavioral data, such as the city the user is in, and the specific indoor or outdoor environment, allows analysis of the impact of different locations on health data and device usage.

[0052] Activity zone labeling: Label the user's daily activity zone (such as home, office, sports venue, etc.) to analyze health status and equipment requirements in different activity zones. For example, the equipment requirements in the office and at home may be significantly different.

[0053] Data denoising: Remove abnormal or incomplete data, such as missing health parameters or misoperated equipment usage records, to ensure data quality.

[0054] Assume that user A is a 50-year-old office worker who has installed multiple health monitoring devices, such as a smart bracelet and a home blood pressure monitor. The user usually pays attention to blood pressure, heart rate, and sleep quality, and has some basic care equipment at home.

[0055] 1) Behavioral data collection

[0056] Health parameters: The system collects user A’s health data from the smart bracelet, including:

[0057] Blood pressure: Measure once every morning and evening.

[0058] Heart rate: Monitor 24 / 7, especially during work and sleep.

[0059] Sleep Quality: Records the duration and depth of sleep during the night.

[0060] Equipment usage records: record the frequency and duration of user A's use of existing nursing equipment, for example:

[0061] Home blood pressure monitor: Use every morning and evening, for about 2 minutes each time.

[0062] Sleep aids: Occasionally used at night.

[0063] 2) Time tag addition

[0064] Intraday time period label: Label the data according to the time period of the day, such as "Morning" (6-9 am), "Daytime" (9-6 pm), "Evening" (18-22 pm), and "Night" (22-6 pm).

[0065] Weekly periodicity labeling: The system labels data as weekdays or weekends to identify behavioral changes during different periods.

[0066] Special Event Tags: Record specific tags during special events, such as periods of high stress or recovery from illness, to focus on changes in your health during these times.

[0067] 3) Spatial label addition

[0068] Geolocation tagging: The system detects the user's geolocation changes between home and work and analyzes data in different environments.

[0069] Activity zone labels: Label the zone data of your home and office to facilitate analysis of the impact of different environments on user health.

[0070] Use the K-means algorithm to perform cluster analysis on user behavior data to identify typical user behavior patterns, such as frequency of getting up at night and amount of exercise. Based on the clustering results, users are divided into different groups so that the behavioral characteristics of similar groups can be considered when making recommendations, thereby improving the accuracy and applicability of recommendations. The following steps are included:

[0071] Randomly select K data points as the initial cluster centers, calculate the distance between each data point and the cluster center, assign the data point to the cluster with the smallest distance, recalculate the cluster center for each cluster, and set it as the mean of all data points in the cluster. When the cluster center no longer changes or reaches the preset number of iterations, the algorithm converges.

[0072] Based on the clustering results, analyze the characteristic values ​​of each cluster center (such as frequency of getting up at night, amount of exercise, etc.) and extract typical behavior patterns. For example:

[0073] Cluster 1: High frequency of getting up at night and low amount of exercise.

[0074] Cluster 2: Low frequency of getting up at night and high amount of exercise.

[0075] Cluster 3: Moderate frequency of nighttime waking and amount of exercise.

[0076] Name and explain clusters: Name each cluster group (such as "high activity and low sleep group", "healthy regularity group", etc.) to facilitate the application of these grouping labels in recommendations.

[0077] Based on the clustering results, each user is assigned to the most similar group. For example, user A may be assigned to the "high activity and low sleep group". The number of users and feature distribution of each group are calculated to obtain the proportion and feature distribution of different groups in the overall user population.

[0078] Suppose we have a batch of user behavior data, including:

[0079] Frequency of waking up at night: indicates the average number of times a user wakes up in one night.

[0080] Weekly exercise volume: The amount of exercise a user does in a week is expressed in hours.

[0081] Average Nighttime Heart Rate: represents the average heart rate of the user at night.

[0082] Daily steps: indicates the user's average daily steps.

[0083] K=3 is selected, that is, users are divided into 3 groups. The system randomly selects 3 data points from the user data set as the initial cluster centers, for example:

[0084] Initial cluster center 1: high frequency of waking up, low amount of exercise, medium heart rate at night, and low number of steps;

[0085] Initial cluster center 2: low frequency of waking up, high amount of exercise, low heart rate at night, and high number of steps;

[0086] Initial cluster center 3: The frequency of waking up and the amount of exercise are moderate, and the heart rate and steps at night are both at moderate levels.

[0087] Calculate the distance between each user and the centers of the three initial clusters, and assign the user to the cluster with the closest distance. For example:

[0088] User A: gets up twice at night, exercises for 5 hours a week, has a nighttime heart rate of 65 beats / minute, and takes 6,000 steps per day. He is closest to cluster center 2 and is therefore assigned to cluster 2.

[0089] User B: gets up 4 times at night, exercises for 2 hours a week, has a nighttime heart rate of 80 beats / minute, and takes 3,000 steps per day on average. He is closest to cluster center 1 and is therefore assigned to cluster 1.

[0090] After all users are assigned, recalculate the center of each cluster and set it to the mean of all user features in the cluster. For example:

[0091] The new center of cluster 1 may be updated as: waking up 3.5 times, exercising 2.5 hours per week, nighttime heart rate 78 beats / minute, and average daily steps 3,200 steps.

[0092] The new center of cluster 2 may be updated as: waking up frequency 1.2 times, exercising 6 hours per week, nighttime heart rate 64 beats / minute, and average daily steps 7200 steps.

[0093] The data allocation and cluster center update process is repeated until the cluster center no longer changes or the preset number of iterations is reached.

[0094] After the algorithm converges, we can get three typical user behavior patterns:

[0095] Cluster 1 (high frequency of nighttime waking and low physical activity):

[0096] Getting up 3-4 times at night, low weekly exercise volume (2-3 hours), high heart rate at night (about 78 beats / minute), and low number of steps.

[0097] Named "high wake-up and low exercise group".

[0098] Cluster 2 (low frequency of nighttime waking and high physical activity):

[0099] Get up less than 2 times at night, have a higher weekly exercise volume (5-7 hours), a low heart rate at night (about 65 beats / minute), and a high number of steps.

[0100] Named "Healthy Regularity Group".

[0101] Cluster 3 (moderate frequency of nighttime waking and moderate physical activity):

[0102] I get up 2-3 times at night, exercise about 4 hours a week, and my heart rate and steps at night are moderate.

[0103] Named "Moderate Activity Group".

[0104] Assign users to groups: Based on the clustering results, assign users to the most similar groups. For example:

[0105] User A is assigned to the "healthy regularity group" and user B is assigned to the "high waking up and low exercise group".

[0106] Group characteristics distribution and proportion analysis:

[0107] The “healthy routine group” has 40% of users, a high proportion, showing healthy lifestyle patterns.

[0108] The "high waking up and low exercise group" has 30% of users, and it is recommended to increase the recommendation of equipment for sleep and exercise management.

[0109] The “moderate activity group” has 30% of users, who are suitable for basic health monitoring device recommendations.

[0110] The semivariance function is used to describe the similarity of user behavior data to form a semivariance model, which includes the following steps:

[0111] In the user behavior data, select the feature pairs that need to describe similarity (such as the frequency of getting up at night and the amount of exercise). For each pair of user data points, calculate the difference between the feature values. The expression is: ΔX = X i -X j , where ΔX is the difference, X i is the data point of user i, X j is the data point of user j, and the square of the difference is used to get the variation value, which is expressed as: In the formula, r is the variation value;

[0112] The user location data is divided into different distance intervals (for example, 1 km, 2 km, etc.) to analyze the relationship between distance and variation. The user data pairs with close distances are grouped according to the distance interval. Each group contains all the user pairs in the distance interval. For each distance interval, the mean square variation value of all data pairs in it is calculated as the semi-variance value of the distance to complete the construction of the semi-variance model. The semi-variance model expression is:

[0113] Where R is the semivariance value, N(h) represents the number of data pairs in the distance interval, and the semivariance values ​​at different distances are plotted to obtain the semivariogram, which is used to describe the relationship between distance and variation.

[0114] After selecting the corresponding gap filling model based on the semivariance model feature, the gap filling model includes a common gap filling model, a general gap filling model, etc. The gap filling model is selected based on the clustered user behavior data to estimate the behavior data missing in time or space. The interpolation result can fill the blank area of ​​the user behavior history record and generate more complete behavior data, including the following steps:

[0115] in;

[0116] Ordinary gap filling model: suitable for situations where the data mean is relatively stable;

[0117] Universal gap filling model: It is suitable for situations where the data mean changes with spatial position and where the behavioral characteristics have significant differences in different locations or time periods.

[0118] Based on the clustering results, the data of a specific cluster is extracted as a training set, and the gap filling model is trained on the selected cluster data to determine the model parameters so that the interpolation effect is optimal. The model parameters are optimized through methods such as cross-validation to improve the accuracy of interpolation and ensure that the interpolation results can reasonably reflect the behavior pattern of the cluster center.

[0119] The location of the behavior data missing in time or space is estimated by the gap filling model, and the gap filling model expression is: In the formula, is the predicted value at position s0, that is, the missing behavior data value we want to estimate, X(s i ) is the position s i The observed value of the known data point at position s i The behavioral data value at (e.g., frequency of getting up at night, amount of exercise, etc.), N represents the total number of known data points used for interpolation, and λ i Indicates position s i The interpolation weight for position s0. Fill the blank area of ​​the user behavior data with the interpolation result to generate a complete history record.

[0120] In this application, the interpolation weight λ i The acquisition logic is: Get location s i The observed value fluctuation amplitude and position s i The semivariance value and semivariance sill value between the position s0 are normalized to map the value range of the observation value fluctuation amplitude, semivariance value and semivariance sill value to [0,1]. The observation value fluctuation amplitude, semivariance value and semivariance sill value after normalization are summed to obtain the position s i The sum value is taken as the inverse of the sum value as the influence value. The larger the influence value, the higher the position s i The greater the impact on position s0, the sum of the impact values ​​of all positions is used to obtain the impact denominator value, and the impact value is divided by the impact denominator value to obtain the impact value of each position s i The interpolation weight λ i .

[0121] Specific:

[0122] The smaller the semivariance, the higher the correlation of the behavioral data between locations, and the larger the corresponding weight. The semivariance sill value represents the overall variability of the data and determines the level at which the semivariance reaches its maximum value. The sill determines the upper limit of the overall correlation of the data at different locations. A larger sill means that the data has a greater overall variability and has less impact on distant locations; vice versa.

[0123] The calculation expression of the observed value fluctuation amplitude is: Where σ is the fluctuation amplitude of the observed value, M is the number of observed values, and Z(s i ) indicates position s i The i-th observation value of Z(s i ) is the mean of the observed values;

[0124] Large fluctuations in observed values ​​may reflect irregularities in user behavior, and using such data may introduce greater uncertainty. Data with greater volatility can be given a smaller weight to ensure that the interpolation results are more stable.

[0125] Combine the characteristic data of user behavior data (such as sleep quality, daytime activity, device usage frequency, etc.) and the prediction results of the gap filling model, assign weights to each characteristic data, use the multi-factor weighted method to set recommendation rules, and generate a nursing device recommendation list based on the recommendation rules. The list is displayed in order of priority, including the following steps:

[0126] Description: Set recommendation rules based on weighted feature data and the results of the gap filling model. You can define rules such as:

[0127] Criteria for preferred equipment:

[0128] Frequently used devices: Recommend devices that are frequently used by users.

[0129] Improve sleep quality: Prioritize devices that can improve sleep quality.

[0130] Increase daily activity: Recommend devices that motivate users to increase their daily activity.

[0131] According to the recommendation rules, a recommendation list is generated based on the feature score of each device, and the comprehensive score of each nursing device is calculated: In the formula, S i is the comprehensive score of the i-th nursing device, m is the number of feature data, and w j is the weight of the jth feature data, F ij For the performance of the i-th nursing device on the j-th feature data, all nursing devices are sorted from large to small according to the comprehensive score to generate a list of recommended devices.

[0132] Specifically, the performance of the i-th nursing device on the j-th feature data is obtained based on user feedback.

[0133] The recommended nursing devices are displayed to users according to priority, including information such as the characteristics of each device, applicable population and usage effects.

[0134] The interface design can be displayed in card format, and the information of each device includes:

[0135] Device name;

[0136] Comprehensive score;

[0137] Characteristic data (such as improved sleep quality, increased activity, etc.);

[0138] Frequency of use and user feedback;

[0139] Through the above steps, the user's behavioral data characteristics and the prediction results of the gap-filling model can be combined to generate a personalized nursing equipment recommendation list, and the list is displayed in order of priority, thereby improving the accuracy of the recommendation and user satisfaction.

[0140] Suppose we have the following user behavior data and device data:

[0141] User behavior data:

[0142] User A:

[0143] Sleep quality: 6 hours (30% deep sleep);

[0144] Daily activity: 3,000 steps, 2 hours of activity time;

[0145] Frequency of equipment use: Nursing equipment is used 3 times / week;

[0146] Nursing equipment data:

[0147] Device 1: Smart sleep monitor;

[0148] Improve sleep quality: 8 points;

[0149] Increased daytime activity: 4 points;

[0150] Frequency of equipment use: 6 points;

[0151] Device 2: Smart sports bracelet:

[0152] Improve sleep quality: 4 points;

[0153] Increased daytime activity: 9 points;

[0154] Frequency of device use: 5 points;

[0155] Device 3: Intelligent health management platform:

[0156] Increased daytime activity: 5 points;

[0157] Device usage frequency: 7 points.

[0158] Assume that the weight of each feature is determined through statistical analysis: the weight of improving sleep quality is equal to 0.5, the weight of increasing daytime activity is equal to 0.3, and the weight of device use frequency is equal to 0.2;

[0159] Calculate the overall score for each nursing device: S i =w1·F i1 +w2·F i2 +w3·F i3 Where, F i1 、F i2 、Fi3 are the performances of the ith nursing device in improving sleep quality, increasing daytime activity, and device usage frequency. w1, w2, and w3 are the weights of improving sleep quality, increasing daytime activity, and device usage frequency, respectively.

[0160] After calculation, the comprehensive score of device 1: smart sleep monitor is 6.4, the comprehensive score of device 2: smart sports bracelet is 5.7, and the comprehensive score of device 3: smart health management platform is 5.9;

[0161] The recommended equipment list is sorted as follows:

[0162] 1) Smart sleep monitor (6.4 points);

[0163] 2) Intelligent health management platform (5.9 points);

[0164] 3) Smart sports bracelet (5.7 points).

[0165] Embodiment 3: The nursing equipment recommendation system based on user historical behavior described in this embodiment includes a data collection module, a clustering module, a model training module, and a recommendation module;

[0166] Data collection module: collects user behavior data, including health parameters (such as blood pressure, heart rate, body temperature, etc.) and the frequency and duration of using nursing equipment, adds time and space tags to the behavior data (such as time of day, geographic location, etc.), and sends the behavior data to the clustering module and model training module;

[0167] Clustering module: Use the K-means algorithm to perform cluster analysis on user behavior data, identify typical user behavior patterns, such as frequency of getting up at night, amount of exercise, etc., and divide users into different groups based on the clustering results, so that the behavioral characteristics of similar groups can be considered when making recommendations, thereby improving the accuracy and applicability of recommendations. The clustering results are sent to the model training module;

[0168] Model training module: Use the semivariance function to describe the similarity of user behavior data to form a semivariance model. After selecting the corresponding gap filling model based on the semivariance model features, the gap filling model includes a common gap filling model, a general gap filling model, etc. The selected gap filling model is trained based on the clustered user behavior data to estimate the behavior data missing in time or space. The interpolation result can fill the blank area of ​​the user behavior history record and generate more complete behavior data. The interpolation result is sent to the recommendation module;

[0169] Recommendation module: Combine the characteristic data of user behavior data (such as sleep quality, daytime activity, frequency of device use, etc.) and the prediction results of the gap-filling model, assign weights to each characteristic data, use the multi-factor weighted method to set recommendation rules, and generate a recommended list of nursing equipment based on the recommendation rules. The list is displayed in order of priority.

[0170] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0171] In the description of this specification, the description with reference to the terms "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0172] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to only specific implementation methods. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A nursing device recommendation method based on user historical behavior, characterized in that: The recommended method comprises the following steps: Collect user behavior data, add time and space labels to the behavior data, use the K-means algorithm to perform cluster analysis on the user behavior data, identify user behavior patterns, and divide users into different groups based on the clustering results; The semivariance function is used to describe the similarity of user behavior data to form a semivariance model. After selecting the corresponding gap filling model based on the semivariance model features, the selected gap filling model is trained based on the clustered user behavior data to estimate the behavior data missing in time or space. Combine the characteristic data of user behavior data and the prediction results of the gap-filling model, assign weights to each characteristic data, use the multi-factor weighted method to set recommendation rules, and generate a nursing equipment recommendation list based on the recommendation rules.

2. The nursing device recommendation method based on user historical behavior according to claim 1 is characterized in that: Using the semivariance function to describe the similarity of user behavior data to form a semivariance model includes the following steps: In the user behavior data, select the feature pairs that need to describe the similarity. For each pair of user data points, calculate the difference between the feature values. The expression is: ΔX = X i -X j , where ΔX is the difference, X i is the data point of user i, X j is the data point of user j, and the square of the difference is used to get the variation value, which is expressed as: In the formula, r is the variation value; The user location data is divided into different distance intervals, and the user data pairs are grouped according to the distance interval. Each group contains all the user pairs in the distance interval. The mean square variation value of all data pairs in each distance interval is calculated as the semivariance value of the distance to complete the construction of the semivariance model. The semivariance model expression is: Where R is the semivariance value, N(h) represents the number of data pairs in the distance interval, and the semivariance values ​​at different distances are plotted to obtain the semivariogram, which is used to describe the relationship between distance and variation.

3. The nursing device recommendation method based on user historical behavior according to claim 2 is characterized in that: After selecting the corresponding gap filling model based on the semivariance model feature, the selected gap filling model is trained according to the clustered user behavior data to estimate the behavior data missing in time or space, including the following steps: Based on the clustering results, the cluster data is extracted as the training set, and the gap filling model is trained on the selected cluster data; The location of the behavior data missing in time or space is estimated by the gap filling model, and the gap filling model expression is: In the formula, is the predicted value at position s0, that is, the missing behavior data value to be estimated, X(s i ) is the position s i The observation value of the known data point at position s i The behavior data value at, N represents the total number of known data points used for interpolation, λ i Indicates position s i For the interpolation weight of position s0, the interpolation result is filled into the blank area of ​​the user behavior data to generate a complete historical record.

4. The nursing device recommendation method based on user historical behavior according to claim 3 is characterized in that: The interpolation weight λ i The acquisition logic is: Get location s i The observed value fluctuation amplitude and position s i The semivariance value and semivariance sill value between the position s0 are normalized to map the value range of the observation value fluctuation amplitude, semivariance value and semivariance sill value to [0,1]. The observation value fluctuation amplitude, semivariance value and semivariance sill value after normalization are summed to obtain the position s i Sum the value, take the inverse of the sum as the influence value, sum the influence values ​​of all positions to get the influence denominator value, and divide the influence value by the influence denominator value to get the influence value of each position s i The interpolation weight λ i .

5. The nursing device recommendation method based on user historical behavior according to claim 4 is characterized in that: Use the K-means algorithm to perform cluster analysis on user behavior data and identify user behavior patterns, including the following steps: Randomly select K data points as the initial cluster centers, calculate the distance between each data point and the cluster center, and assign the data point to the cluster with the smallest distance; Recalculate the cluster center for each cluster and set the cluster center to the mean of all data points in the cluster. When the cluster center no longer changes or reaches the preset number of iterations, the algorithm converges. According to the clustering results, analyze the characteristic values ​​of each cluster center, extract typical behavior patterns, and name each cluster group; According to the clustering results, each user is assigned to a similar group, and the number of users and feature distribution of each group are calculated to obtain the proportion and feature distribution of different groups in the overall users.

6. The nursing device recommendation method based on user historical behavior according to claim 5 is characterized in that: Combine the characteristic data of user behavior data and the prediction results of the gap filling model, assign weights to each characteristic data, use a multi-factor weighted method to set recommendation rules, and generate a nursing equipment recommendation list based on the recommendation rules, including the following steps: Set recommendation rules based on weighted feature data and the results of the gap-filling model; According to the recommendation rules, a recommendation list is generated based on the feature score of each device, and the comprehensive score of each nursing device is calculated: In the formula, S i is the comprehensive score of the i-th nursing device, m is the number of feature data, w j is the weight of the jth feature data, F ij For the performance of the i-th nursing device on the j-th feature data, all nursing devices are sorted from large to small according to the comprehensive score to generate a list of recommended devices; The recommended nursing devices are displayed to users according to priority, including the characteristics of each device, applicable population and usage effect information.

7. The nursing device recommendation method based on user historical behavior according to claim 6 is characterized in that: The recommendation system collects user behavior data and adds time and space labels to the behavior data, including the following steps: Through smart wearable devices and smart monitoring devices, users' health data, including blood pressure, heart rate, body temperature, sleep quality, and activity level, are collected in real time, and the frequency and duration of users' use of various care devices are recorded; Add a label of the user's time period to each piece of behavior data to analyze the user's health changes and device usage habits in different time periods, and mark the date of the behavior data to capture the periodic changes in the behavior data; Geographic location information is added to behavioral data to analyze the impact of different locations on health data and device usage, and activity area labels are added to users to analyze health status and device requirements in different activity areas.

8. A nursing device recommendation system based on user historical behavior, used to implement the recommendation method according to any one of claims 1 to 7, characterized in that: Includes data collection module, clustering module, model training module, and recommendation module; Data collection module: collects user behavior data and adds time and space tags to the behavior data; Clustering module: Use K-means algorithm to perform cluster analysis on user behavior data, identify user behavior patterns, and divide users into different groups based on clustering results; Model training module: Use the semivariance function to describe the similarity of user behavior data to form a semivariance model. After selecting the corresponding gap filling model based on the semivariance model features, the selected gap filling model is trained based on the clustered user behavior data to estimate the behavior data missing in time or space. Recommendation module: Combine the characteristic data of user behavior data and the prediction results of the gap-filling model, assign weights to each characteristic data, use the multi-factor weighted method to set recommendation rules, and generate a recommended list of nursing equipment based on the recommendation rules.

Images