Animal Behavior Management System Based on Intelligent Dog Training Device

Abstract

The present invention relates to the technical field of information retrieval, specifically an animal behavior management system based on an intelligent dog trainer. The animal behavior management system based on the intelligent dog trainer includes a behavior data acquisition module, a high-dimensional offset calculation module, a behavior boundary recognition module, an abnormal impact degree analysis module, and a behavior index update module. In the present invention, by integrating environmental factors and multi-dimensional behavior data, the comprehensiveness and accuracy of behavior monitoring are improved. By using time series analysis and comprehensive processing of motion data, abnormal behaviors are accurately identified and predicted, enhancing the accuracy of behavior pattern recognition. By analyzing the changes in motion rhythm and the transfer of behavior paths, abnormal behaviors are stably identified, improving the system's ability to handle complexity. Dynamically updating the behavior index and optimizing the matching degree enable the system to adapt to the evolution of behaviors, improving the rationality and effectiveness of intervention strategies.

Description

Technical Field

[0001] The present invention relates to the technical field of information retrieval, and particularly to an animal behavior management system based on an intelligent dog trainer. Background Art

[0002] The technical field of information retrieval includes methods for data storage, organization, query, and retrieval, and is widely applied in application scenarios such as database management, search engines, intelligent recommendation systems, etc. The core content of this technical field involves constructing data index structures, designing query optimization strategies, improving data retrieval efficiency, etc., and usually adopts inverted indexes, Boolean retrieval, vector space models, etc. to improve data matching degree and retrieval speed. In addition, this field also covers technologies such as natural language processing, semantic analysis, and machine learning to enhance the relevance and intelligence of information, facilitating users to quickly obtain the required information from a large amount of data.

[0003] Among them, an animal behavior management system based on an intelligent dog trainer refers to using information retrieval technology to collect, analyze, store, and query animal behavior data to achieve the management of animal behavior. This system collects animal activity data through animal behavior sensing devices, analyzes behavior patterns based on feature extraction methods, and stores the data in a structured database. Subsequently, a behavior information index is constructed through data index technology to achieve retrieval queries based on keywords, time series, and behavior patterns, and support multi-dimensional data analysis. This system also uses classification methods to classify behavior data for quickly retrieving specific behavior records, combines pattern matching methods to identify abnormal behaviors, and implements automatic behavior intervention strategies based on rule settings.

[0004] The prior art is limited to single behavior characteristics in data collection and fails to fully consider environmental factors, resulting in the susceptibility of behavior recognition to interference in complex environments and reducing the accuracy of behavior data. In terms of abnormal behavior analysis, it mainly relies on fixed pattern matching and lacks the analysis of behavior transfer trends, making it difficult to effectively distinguish short-term abnormalities from long-term abnormalities, resulting in unstable judgment of abnormal behaviors. In terms of behavior data storage and indexing, it only relies on static retrieval methods and fails to establish dynamic behavior associations, making it difficult to track changes in behavior patterns, resulting in insufficient adaptability for long-term behavior management. In terms of the impact analysis of abnormal behaviors, the evaluation method is too static and lacks quantitative analysis of the impact range of abnormal behaviors, resulting in the lack of pertinence of abnormal intervention strategies, which may cause over-intervention or under-intervention and reduce the effectiveness of intervention measures. Summary of the Invention

[0005] The purpose of the present invention is to solve the drawbacks existing in the prior art, and to propose an animal behavior management system based on an intelligent dog trainer.

[0006] To achieve the above purpose, the present invention adopts the following technical solution: The animal behavior management system based on an intelligent dog trainer includes:

[0007] The behavior data acquisition module obtains the behavior data of the dog, calls the sensor data of the dog trainer, records the movement trajectory and environmental factors, and screens the environmental behavior feature data.

[0008] The high-dimensional offset calculation module extracts time series features based on the environmental behavior feature data, calculates the trajectory position difference and offset rate, records the current behavior path, judges the change rate of the offset, and combines the movement acceleration and direction change to screen the behavior transition rate, generating a dynamic offset prediction result.

[0009] The behavior boundary recognition module calls the dynamic offset prediction result, records the continuity break of abnormal behaviors, divides time periods, calculates the movement frequency, sprint ratio, and stationary time, analyzes the change of movement rhythm, identifies the deviation of the behavior path, and generates a behavior fuzzy interval.

[0010] The abnormal impact analysis module evaluates the impact of abnormal behaviors on the normal mode based on the behavior fuzzy interval, calculates the decrease in behavior transfer probability, behavior delay, and break rate, and generates the impact ratio of abnormal behaviors.

[0011] The behavior index update module calls the impact ratio of abnormal behaviors, analyzes the change of behavior intervals, calculates the matching degree of new behaviors, compares the probability of repeated behaviors, screens the index update data, and generates an intervention strategy after behavior update.

[0012] As a further solution of the present invention, the environmental behavior feature data includes movement feature data, trajectory data, and environmental feature data; the dynamic offset prediction result includes trajectory change features, offset rate features, and abnormal behavior offset values; the behavior fuzzy interval includes behavior continuity features, time partition features, and behavior path change features; the impact ratio of abnormal behaviors includes behavior transfer features, abnormal behavior impact features, and behavior break features; and the intervention strategy after behavior update includes behavior interval features, new behavior matching features, and behavior index update features.

[0013] As a further solution of the present invention, the behavior data acquisition module includes:

[0014] The movement data extraction sub-module calls the built-in sensor data of the intelligent dog trainer to obtain the movement acceleration, behavior sequence, and activity frequency of the dog, calculates the behavior duration, statistically analyzes the rhythm change rate and the stability of adjacent behavior sequences, analyzes the movement trajectory points of the dog, records the number of sprints and the ratio of stationary time, and generates movement feature data;

[0015] The behavior pattern analysis sub-module identifies the continuous behavior patterns of the dog based on the movement feature data, analyzes the behavior transition trend, extracts the occurrence probability of behaviors, calculates the correlation degree between different behaviors of the dog, and uses the formula:

[0016] ;

[0017] Calculate the stability coefficient of the dog's behavior sequence through operations, and combine the behavior conversion trend to obtain the behavior pattern characteristics;

[0018] Among them, represents the behavior pattern characteristics, represents the duration of a single behavior, represents the behavior conversion interval, is the total number of behaviors, is the time decay coefficient, represents the total duration of the behavior sequence, represents the sum of the products of the durations of all behaviors and the behavior conversion intervals;

[0019] Based on the behavior pattern characteristics, the environmental feature screening sub-module combines the noise level, light intensity, and ground vibration conditions in the space where the dog is located, calculates the impact of environmental feature changes on the behavior pattern, screens highly correlated environmental variables, and obtains environmental behavior feature data.

[0020] As a further solution of the present invention, the high-dimensional offset calculation module includes:

[0021] Based on the environmental behavior feature data, the trajectory position calculation sub-module extracts the time series behavior features, obtains the motion trajectory coordinates of adjacent time points, calculates the trajectory position change value within each time step, sets the time interval, calculates the trajectory vector change amount, and combines the trajectory data to obtain the trajectory position difference;

[0022] Based on the trajectory position difference, the offset rate analysis sub-module calculates the offset rate between adjacent behavior feature vectors, using the formula:

[0023] ;

[0024] Calculate the behavior offset rate through operations, compare the behavior trajectory data with the current behavior path data, and obtain the change rate of the behavior offset amount;

[0025] Among them, represents the behavior offset rate, represents the th trajectory coordinate value of the time point, represents the th target trajectory coordinate value of the time point, is the total number of trajectory points within the time step, is the trajectory offset adjustment factor, represents the offset trend of the behavior trajectory, is the correction amount of the behavior vector change rate;

[0026] Based on the rate of change of the behavior offset, the abnormal behavior screening sub-module calculates the abnormal behavior offset value, combines the acceleration and direction changes of the dog's movement, screens the data with a behavior transition rate exceeding 0.8, and obtains the deviation dynamic prediction result.

[0027] As a further solution of the present invention, the behavior boundary recognition module includes:

[0028] Based on the deviation dynamic prediction result, the behavior continuity calculation sub-module extracts the behavior sequences before and after the abnormal behavior, calculates the time intervals between consecutive behaviors, counts the behavior disruption rate, divides the behavior time periods, and obtains the behavior continuity disruption value;

[0029] Based on the behavior continuity disruption value, the movement rhythm analysis sub-module calculates the movement frequency, sprint proportion, and stationary duration for each time period, constructs a behavior change matrix within the time period using the multi-scale behavior distribution analysis method, analyzes the stability of the movement state in the time partition, and uses the formula:

[0030] ;

[0031] Performs calculations to obtain the movement rhythm change amplitude, compares the calculation results of adjacent time periods, constructs a behavior transfer trend vector, establishes a time dynamic transfer matrix, and obtains the behavior transfer change coefficient;

[0032] Wherein, represents the movement rhythm change amplitude, represents the movement frequency value at the th time point, represents the stationary duration at the th time point, is the total number of behavior records within the time step, is the behavior fluctuation weight, represents the behavior change rate at the th time point, represents the behavior change period parameter, is the behavior change smoothing factor, represents the behavior interval fluctuation degree, is the movement adjustment factor, represents the sprint behavior proportion change parameter, represents the dispersion correction parameter of the behavior rhythm within the time period;

[0033] Based on the behavior transfer change coefficient, the behavior path comparison sub-module compares the normal behavior path and the abnormal behavior path, calculates the path fitting degree index, screens the similar path intervals, and obtains the behavior fuzzy interval.

[0034] As a further aspect of the present invention, the abnormal impact degree analysis module includes:

[0035] The behavior sequence stability evaluation sub-module extracts the behavior sequences before and after the abnormal behavior based on the behavior fuzzy interval, compares the behavior stability in the time periods before and after the abnormality, calculates the distribution probability of behavior categories within the time period, calculates the behavior frequency fluctuation, determines the perturbation degree of the behavior sequence, and calculates the behavior stability decline ratio according to the perturbation measurement standard to obtain the behavior sequence stability offset value;

[0036] The behavior transfer impact calculation sub-module calculates the decline amplitude of the behavior transfer probability within the abnormal behavior time period based on the behavior sequence stability offset value, calculates the conversion rate from normal behavior to abnormal behavior, constructs a time-segmented behavior impact model, calculates the dynamic impact of the abnormal behavior on the normal behavior sequence, and uses the formula:

[0037] ;

[0038] Performs operations to obtain the behavior transfer fracture rate, analyzes the continuous interference effect of abnormal behavior on normal behavior in the behavior pattern, and calculates the overall change amplitude of the behavior pattern.

[0039] Wherein, represents the behavior transfer fracture rate, represents the normal behavior transfer probability at the th time point, represents the transfer probability when the abnormal behavior occurs, is the total number of behavior records within the time step, represents the duration of normal behavior, represents the duration of abnormal behavior;

[0040] The abnormal behavior impact evaluation sub-module calculates the impact proportion of the abnormal behavior on the normal behavior pattern based on the behavior transfer fracture rate, extracts the impact of the abnormal behavior on the normal behavior, analyzes the expansion range of the abnormal behavior on the time axis, analyzes the impact degree of the abnormal behavior on the behavior stability, and establishes a measure index for the behavior pattern affected by the abnormality to obtain the abnormal behavior impact proportion.

[0041] As a further aspect of the present invention, the behavior index update module includes:

[0042] The behavior interval analysis sub-module calls the abnormal behavior impact proportion to obtain the duration sequence of the dog's behavior, calculates the behavior interval time based on the behavior data, calculates the behavior interval change rate, analyzes the correlation between the recurrence probability of abnormal behavior and the normal behavior interval, and extracts the behavior interval change trend, using the formula:

[0043] ;

[0044] Calculate the change ratio of behavior intervals , and obtain the change trend of behavior intervals;

[0045] Among them, represents the change ratio of behavior intervals, is the interval time of the th behavior, is the interval time of the previous behavior, is the total number of times the behavior occurs;

[0046] Based on the change trend of behavior intervals, the behavior matching degree evaluation sub-module constructs a behavior persistence weight matrix, calculates the matching degree of abnormal behaviors in the overall behavior pattern, evaluates the matching degree, compares the recurrence probability of normal behaviors, screens candidate behavior data that conform to the behavior pattern, and obtains the behavior matching evaluation result;

[0047] The behavior index screening sub-module calls the behavior matching evaluation result, filters the data nodes of the behavior index, calculates the influence range of the behavior after the index update, compares the proportion of behavior changes before and after the abnormal behavior update, screens the behavior index update scheme, and obtains the intervention strategy after the behavior update.

[0048] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0049] In the present invention, by integrating environmental factors and multi-dimensional behavior data, the comprehensiveness and accuracy of behavior monitoring are improved. By using the comprehensive processing of time series analysis and motion data, abnormal behaviors are accurately identified and predicted, the accuracy of behavior pattern recognition is enhanced. By analyzing the change of motion rhythm and the transfer of behavior paths, abnormal behaviors are stably identified, the system's ability to handle complexity is improved. By dynamically updating the behavior index and optimizing the matching degree, the system can adapt to the evolution of behaviors, and the rationality and effectiveness of the intervention strategy are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 is the system flow chart of the present invention;

[0051] Figure 2 is the flow chart of the behavior data acquisition module of the present invention;

[0052] Figure 3 is the flow chart of the high-dimensional offset calculation module of the present invention;

[0053] Figure 4 is the flow chart of the behavior boundary recognition module of the present invention;

[0054] Figure 5 is the flow chart of the abnormal influence degree analysis module of the present invention;

[0055] Figure 6This is the flowchart of the behavior index update module of the present invention. Detailed implementation manners

[0056] In order to make the objectives, technical solutions and advantages of the present invention more clear and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0057] In the description of the present invention, it should be understood that the orientation or positional relationship indicated by the terms "length", "width", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. is based on the orientation or positional relationship shown in the accompanying drawings, and is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus cannot be understood as a limitation to the present invention. In addition, in the description of the present invention, the meaning of "a plurality of" is two or more, unless otherwise specifically defined.

[0058] Please refer to Figure 1 , the animal behavior management system based on the intelligent dog trainer includes:

[0059] The behavior data acquisition module obtains the behavior data stream of the dog wearing the intelligent dog trainer, collects the motion acceleration, behavior sequence, activity frequency, behavior duration, rhythm change rate, and adjacent behavior order stability, calls the built-in sensing data of the intelligent dog trainer, records the dog's motion trajectory points, sprint times, and the proportion of stationary time, and combines the noise level, light intensity, and ground vibration conditions in the space where the dog is located to screen the environmental behavior feature data;

[0060] The high-dimensional offset calculation module extracts the time-series behavior features based on the environmental behavior feature data, calculates the difference in trajectory positions between adjacent time points, analyzes the offset rate between adjacent behavior feature vectors, judges the change rate of the behavior offset amount by comparing the behavior trajectory data with the current behavior path data, calculates the abnormal behavior offset value according to the behavior offset rate, and combines the dog's motion acceleration and direction change to screen the behavior transition rate to generate a dynamic offset prediction result;

[0061] The behavior boundary recognition module calls the dynamic offset prediction result, records the disruption of behavior continuity before and after the abnormal behavior, divides the time period and calculates the motion frequency, sprint proportion, and stationary duration of each segment, analyzes the change amplitude of the motion rhythm between adjacent time periods according to the behavior change trend in the time partition, identifies the change coefficient of the behavior transfer path, and compares the difference between the normal behavior path and the abnormal behavior path to generate a behavior fuzzy interval;

[0062] The abnormal impact analysis module calls the behavior fuzzy interval, analyzes the stability of the behavior sequence before and after the abnormal behavior according to the transfer path of the dog's behavior pattern, calculates the decline in the behavior transfer probability, records the delay of the behavior, evaluates the impact of the abnormal behavior on the normal behavior pattern, calculates the behavior transfer break rate according to the impact ratio, and generates the abnormal behavior impact ratio.

[0063] The behavior index update module calls the abnormal behavior impact ratio, calculates the change law of the behavior interval time based on the duration of the dog's behavior and the recurrence probability of the abnormal behavior, analyzes the matching degree of the new behavior according to the behavior persistence weight matrix, compares the recurrence probability of the dog's behavior, filters the behavior index update data nodes, and generates the intervention strategy after behavior update.

[0064] The environmental behavior characteristic data includes motion characteristic data, trajectory data, and environmental characteristic data. The deviation dynamic prediction result includes trajectory change characteristics, deviation rate characteristics, and abnormal behavior deviation values. The behavior fuzzy interval includes behavior continuity characteristics, time partition characteristics, and behavior path change characteristics. The abnormal behavior impact ratio includes behavior transfer characteristics, abnormal behavior impact characteristics, and behavior break characteristics. The intervention strategy after behavior update includes behavior interval characteristics, new behavior matching characteristics, and behavior index update characteristics.

[0065] Please refer to Figure 2 , the behavior data acquisition module includes:

[0066] The motion data extraction sub-module calls the built-in sensor data of the intelligent dog trainer, obtains the motion acceleration, behavior sequence, and activity frequency of the dog, calculates the behavior duration, statistically analyzes the rhythm change rate and the stability of adjacent behavior order, analyzes the motion trajectory points of the dog, records the number of sprints and the proportion of stationary time, and generates motion characteristic data.

[0067] First, it is necessary to obtain the real-time data of the accelerometer, gyroscope, and geomagnetic sensor. These data can record the spatial position, movement direction, and acceleration changes of the dog at different time points. For example, assuming a dog is moving with an acceleration of 0.5 m / s², by analyzing the time axis, its movement trajectory can be obtained. Next, the extraction of the behavior sequence depends on the data partitioning of consecutive time windows. When the set time window is 1 s, the motion feature vector within each second can be composed of the triaxial acceleration values (Ax, Ay, Az) and the angular velocity values (Gx, Gy, Gz). Then, data filtering methods are used to eliminate outliers. For example, data points with a noise level exceeding 2 times the standard deviation are regarded as abnormal and eliminated. The calculation of the activity frequency analyzes the dog's gait cycle through Fourier transform. Taking a 10 s data window as an example, if the main step frequency is calculated to be in the range of 1.5 Hz to 3.5 Hz, then the current activity frequency of the dog can be defined as 2.5 Hz. Subsequently, the behavior duration is obtained by calculating the time span of each independent behavior. For example, if the dog runs continuously for 5 seconds, the duration of this behavior is 5 s. The rate of rhythm change is calculated using the ratio of speed change to time. If the dog's speed increases from 0.5 m / s to 1.5 m / s within 2 seconds, the rate of rhythm change is (1.5 - 0.5) / 2 = 0.5 m / s². The stability of the adjacent behavior sequence can be calculated by statistically analyzing the behavior transition probability matrix. For example, if the dog transitions from walking to running in 50% of the cases, the transition probability is 0.5. Based on the analysis of these parameters for the dog's movement trajectory points, geographical coordinates can be collected through the GPS module, and combined with the timestamp to obtain a continuous trajectory. For example, if the dog moves along (X1, Y1) → (X2, Y2) → (X3, Y3) within 10 s, a movement trajectory is finally formed. The record of the number of sprints depends on the detection of speed peaks. For example, if 3 speed peaks exceeding 2 m / s are detected within 10 s, the number of sprints is 3 times. The proportion of stationary time is obtained by statistically analyzing the proportion of time when the speed is close to 0 m / s. For example, if the dog is stationary for 10 s within 30 s, the proportion of stationary time is 10 / 30 = 33.3%. Finally, the motion feature data is obtained.

[0068] Based on the motion feature data, the behavior pattern analysis sub-module identifies the continuous behavior patterns of the dog, analyzes the behavior transition trend, extracts the occurrence probability of behaviors, and calculates the correlation degree between different behaviors of the dog. The formula is used:

[0069] ;

[0070] Through calculation, the stability coefficient of the dog's behavior sequence is obtained. Combining the behavior transition trend, the behavior pattern characteristics are obtained;

[0071] Among them, represents the behavior pattern characteristics, represents the duration of a single behavior, Represents the behavior transition interval, is the total number of behaviors, is the time decay coefficient, represents the total duration of the behavior sequence, represents the sum of the products of the durations of all behaviors and the behavior transition intervals;

[0072] First, classify the motion data by setting behavior labels (such as stationary, walking, running, jumping). For example, if the acceleration is less than 0.2 m / s² and the speed is less than 0.3 m / s, it is judged as the stationary state. If the acceleration is between 0.3 m / s² - 1.0 m / s² and the speed is between 0.3 m / s - 1.5 m / s, it is judged as the walking state. By constructing a Markov state transition matrix, statistically calculate the transition probabilities between different behaviors. For example, the probability of a dog changing from walking to running is 0.6, and the probability of changing from running to stationary is 0.3. The behavior transition trend can be calculated using the behavior sequence data. For example, in 100 behavior samples, if the proportion of dogs changing from walking to running is 60%, then the transition probability is 0.6. In addition, the occurrence probability of behaviors can be calculated through global behavior statistics. For example, in 500 seconds of data, if a dog is in the walking state for 200 seconds, then the occurrence probability of the walking behavior is 200 / 500 = 0.4. The correlation degree between different behaviors can be calculated using the Pearson correlation coefficient. For example, in 30 groups of data samples, the correlation coefficient of the behavior time between walking and running is 0.75, indicating a strong correlation between the two. Use the formula:

[0073] ;

[0074] Among them, is the duration of the behavior. If a dog completes a behavior within 5 seconds, then , is the behavior transition interval. For example, if another behavior occurs 2 seconds after a certain behavior occurs, then , is the time decay coefficient. If it is set , is the total duration of the behavior sequence. Assuming the duration of this behavior sequence is 100 seconds, then . Substituting these data into the calculation, we can get:

[0075] ;

[0076] Finally, obtain the behavior pattern characteristics. The result shows that the stability of the current dog's behavior pattern is relatively low, and there is a high variability during the behavior transition process.

[0077] Based on the behavior pattern features, the environmental feature screening sub-module combines the noise level, light intensity, and ground vibration conditions in the space where the dog is located, calculates the impact of environmental feature changes on the behavior pattern, screens highly correlated environmental variables, and obtains environmental behavior feature data;

[0078] First, the noise level is detected by a sound sensor, and the unit of measurement is dB. For example, the environmental noise is recorded as 50 dB, 65 dB, and 80 dB at different time periods, where 65 dB is the background noise, and when it exceeds this value, it is judged as a high-noise environment. The light intensity is measured by a light sensor, with the unit of lx (lux). For example, the measured values in the early morning, noon, and evening are 200 lx, 15000 lx, and 500 lx respectively, so noon is a high-light environment. The ground vibration condition is collected by an acceleration sensor. For example, the recorded vibration acceleration values are 0.01 m / s², 0.05 m / s², and 0.2 m / s² respectively. If the vibration acceleration is higher than 0.1 m / s², it is considered that there is significant ground vibration. After obtaining these data, by calculating the correlation between each environmental factor and the behavior pattern features, for example, the correlation coefficient between the noise level and behavior stability is -0.6, indicating that a high-noise environment will reduce the stability of the dog's behavior; the correlation coefficient of the light intensity is 0.4, indicating that a high-light environment may increase the dog's activity frequency; the correlation coefficient between ground vibration and behavior conversion is -0.7, indicating that significant ground vibration will lead to unstable behavior patterns. By screening highly correlated environmental variables, environmental behavior feature data is obtained.

[0079] Please refer to Figure 3 , the high-dimensional offset calculation module includes:

[0080] Based on the environmental behavior feature data, the trajectory position calculation sub-module extracts time-series behavior features, obtains the motion trajectory coordinates of adjacent time points, calculates the trajectory position change value within each time step, sets a time interval, calculates the trajectory vector change amount, and combines the trajectory data to obtain the trajectory position difference;

[0081] When extracting time-series behavior features, first, the motion trajectory data of the dog at different time points needs to be obtained, including three-dimensional coordinate information ( ), and perform time serialization processing. For example, at a certain moment the collected coordinate data is (1.2 m, 3.5 m, 0.8 m), and at the data collected at the moment is (1.4 m, 3.8 m, 0.9 m), then the trajectory position change between the two time points can be calculated as , , , and this data is used to calculate the trajectory offset. When the set time interval is 1 s, the rate of the trajectory offset vector can be expressed as , combined with multi-timepoint data, a complete trajectory sequence can be formed. By statistically analyzing the differences in these trajectory positions, the differences in the trajectory positions of the dog at different time steps can be obtained.

[0082] Based on the differences in trajectory positions, the offset rate analysis sub-module calculates the offset rate between adjacent behavioral feature vectors using the formula:

[0083] ;

[0084] Through calculation, the behavioral offset rate is obtained. By comparing the behavioral trajectory data with the current behavioral path data, the change rate of the behavioral offset amount is obtained;

[0085] where represents the behavioral offset rate, represents the trajectory coordinate value at the th time point, represents the target trajectory coordinate value at the th time point, is the total number of trajectory points within the time step, is the trajectory offset adjustment factor, represents the offset trend of the behavioral trajectory, is the correction amount for the change rate of the behavioral vector;

[0086] When analyzing the offset rate of adjacent behavioral feature vectors, it is necessary to extract the feature vectors at adjacent time points in the behavioral sequence. For example, the motion feature vector collected at time point is ( , , ), and the vector collected at time point is ( , , ). Then, the offset of the adjacent feature vectors can be calculated as , , using the formula:

[0087] ;

[0088] where, let , , , , , , , and the calculation gives:

[0089] ;

[0090] ;

[0091] ;

[0092] This calculation obtains the behavior offset rate. By comparing the behavior trajectory data with the current behavior path data according to this rate, the change rate of the behavior offset can be obtained.

[0093] Based on the change rate of the behavior offset, the abnormal behavior screening sub-module calculates the abnormal behavior offset value. Combining the movement acceleration and direction change of the dog, it screens the data with a behavior transition rate exceeding 0.8 to obtain the deviation dynamic prediction result.

[0094] When calculating the abnormal behavior offset value, it is first necessary to set an offset threshold to determine abnormal behavior. For example, the threshold is set to . When the calculated exceeds this value, it is determined as abnormal behavior. Combining the movement acceleration and direction change of the dog, the behavior transition rate is further calculated. For example, the current acceleration of the dog is , and the acceleration at the previous moment is , then the behavior transition rate can be calculated as . For the data with a behavior transition rate exceeding , it is determined as high-offset behavior data, and finally the deviation dynamic prediction result is obtained.

[0095] Please refer to Figure 4 , the behavior boundary recognition module includes:

[0096] Based on the deviation dynamic prediction result, the behavior continuity calculation sub-module extracts the behavior sequences before and after the abnormal behavior, calculates the time intervals between consecutive behaviors, statistically analyzes the behavior disruption rate, divides the behavior time periods, and obtains the behavior continuity disruption value;

[0097] First, a time window needs to be set to segment the behavior data stream for extracting the behavior patterns in the time periods before and after the abnormality. The length of the time window is set to 5 seconds. In this interval, the movement parameters of the dog are collected, including step frequency, trajectory, speed, direction change, etc. The movement state transitions between adjacent time steps are calculated. For example, at time it is recorded that the dog is in a walking state, enters a running state, comes to an emergency stop, then calculate and The time interval between them is recorded, and the acceleration change trend during this period is recorded. Further analyze the stability of behavior conversion. By counting the number of behavior switches per unit time, the behavior switching frequency is obtained, and the behavior stability deviation is calculated. For example, within 30 seconds, the gait of the dog switches 10 times, and the duration of adjacent behaviors is uneven, indicating that the behavior continuity is impaired within this interval. In addition, using the behavior classification label, it is judged whether there is a drastic jump between the previous and subsequent behaviors. For example, the dog directly enters the jumping state from a stationary state and then enters the sprint state, and the change range of consecutive behavior categories is large, then the degree of behavior continuity disruption is high. Further, the behavior change trend is statistically analyzed, and the abnormal behavior density within this time period is calculated. If the density exceeds the set threshold, this time period is marked as an abnormal time window, and finally the behavior continuity disruption value is obtained.

[0098] Based on the behavior continuity disruption value, the motion rhythm analysis sub-module calculates the motion frequency, sprint ratio, and stationary duration for each time period, constructs a behavior change matrix within the time period using the multi-scale behavior distribution analysis method, and analyzes the motion state stability of the time partition. Using the formula:

[0099] ;

[0100] Calculate the motion rhythm change amplitude through operations, compare the calculation results of adjacent time periods, construct a behavior transfer trend vector, establish a time dynamic transfer matrix, and obtain the behavior transfer change coefficient;

[0101] Among them, represents the motion rhythm change amplitude, represents the th motion frequency value at the time point, represents the th stationary duration at the time point, is the total number of behavior records within the time step, is the behavior fluctuation weight, represents the th behavior change rate at the time point, represents the behavior change cycle parameter, is the behavior change smoothing factor, represents the behavior interval fluctuation degree, is the motion adjustment factor, represents the sprint behavior ratio change parameter, represents the discrete degree correction parameter of the behavior rhythm within the time period;

[0102] Calculate the exercise frequency, sprint ratio, and stationary duration for each time period. Using the method of splitting exercise data, partition the time series. For example, within a 30-second monitoring period, divide it into 6 small intervals of 5 seconds each, and calculate the exercise characteristics within each time period, setting the time step size , calculate the exercise frequency within each time step and the stationary time . Then calculate the overall exercise rhythm fluctuation degree using the formula:

[0103] ;

[0104] where, assume , , , , , , , , , calculate to get:

[0105] ;

[0106] ;

[0107] Finally, through calculation, obtain the change range of the exercise rhythm. Compare the calculation results of adjacent time periods, calculate the change value of the behavior transfer trend, and establish a time dynamic transfer matrix to obtain the behavior transfer change coefficient.

[0108] Based on the behavior transfer change coefficient, the behavior path comparison sub-module compares the normal behavior path and the abnormal behavior path, calculates the path fitting degree index, screens the same type of path intervals, and obtains the behavior fuzzy interval.

[0109] Calculate the trajectory difference between adjacent paths. First, extract the trajectory data of the dog within different time periods to obtain the path coordinate point sequence. For example, within a 5-second time interval, the movement trajectory points of the dog are , , , by calculating the spatial displacement difference between adjacent points, the changing trend of the behavior path is obtained, a path conversion matrix is constructed, the normal behavior path is matched with the abnormal behavior path, and the dynamic offset between the paths is calculated. For example, in the case of normal behavior, the path of the dog presents a smooth curve within 10 seconds, while in the abnormal case, there is a sharp offset in the path curve. Then, the degree of path deviation is calculated, and based on the time series analysis method, it is judged whether the path deviation is continuous. If the offset exceeds the set threshold for multiple consecutive time steps, the time interval is determined as an abnormal path interval. Further, the path similarity is calculated, and the matching error between the paths is calculated by the dynamic time warping method. If the error value exceeds the set threshold of the normal behavior path, the abnormal behavior path interval is screened out, and finally the behavior fuzzy interval is obtained.

[0110] Please refer to Figure 5 , the abnormal impact analysis module includes:

[0111] The behavior sequence stability evaluation sub-module extracts the behavior sequences before and after the abnormal behavior based on the behavior fuzzy interval, compares the behavior stability of the time periods before and after the abnormality, calculates the distribution probability of behavior categories within the time period, calculates the fluctuation of behavior frequencies, judges the degree of perturbation of the behavior sequence, and calculates the ratio of the decrease in behavior stability according to the perturbation measurement standard to obtain the behavior sequence stability offset value;

[0112] First, a time window is constructed to divide the behavior data stream, and the data before and after the occurrence of the abnormal behavior are stored separately and time series analysis is performed. A time step is set to 5 seconds. The distribution of the dog's behavior categories within each time window is counted, including states such as stationary, walking, running, jumping, etc., and the occurrence frequencies of each category within the time window are calculated. For example, the number of behavior category conversions within a 5-second window is 8 times, where the stationary state accounts for 20%, the walking state accounts for 50%, and the running state accounts for 30%. The stability index of the behavior switch is calculated. If the deviation of the behavior category distribution between the front and back windows exceeds 20%, it is determined that the behavior sequence stability has decreased. In addition, the duration of the abnormal behavior is calculated and compared with the average duration of normal behavior . If then the delay situation of the abnormal behavior is recorded, and the influence range of the abnormal behavior on the subsequent behavior stability is counted, and finally the behavior sequence stability offset value is obtained.

[0113] The behavior transfer impact calculation sub-module calculates the decrease amplitude of the behavior transfer probability within the abnormal behavior period based on the behavior sequence stability offset value, calculates the conversion rate from the normal behavior to the abnormal behavior, constructs a time-segmented behavior impact model, and calculates the dynamic impact of the abnormal behavior on the normal behavior sequence, using the formula:

[0114] ;

[0115] Obtain the behavior transfer break rate through calculation, analyze the continuous interference effect of abnormal behaviors on normal behaviors in the behavior pattern, and calculate the overall change amplitude of the behavior pattern.

[0116] Among them, represents the behavior transfer break rate, represents the th normal behavior transfer probability at the time point, represents the transfer probability when an abnormal behavior occurs, is the total number of behavior records within the time step, represents the duration of normal behavior, represents the duration of abnormal behavior;

[0117] Suppose we monitor a dog within a 10-second time window, where the first 5 seconds are normal behaviors and the last 5 seconds are abnormal behaviors. The number of sampling points for normal behaviors is and the number of sampling points for abnormal behaviors is .

[0118] Calculate the change in behavior transfer probability

[0119] Set the transfer probabilities of normal behaviors at 5 time points as:

[0120] ;

[0121] Set the transfer probabilities of abnormal behaviors at 5 time points as:

[0122] ;

[0123] Calculate the change in transfer probability at each time point:

[0124] ;

[0125] ;

[0126] ;

[0127] ;

[0128] ;

[0129] Calculate the average value of the change in behavior transfer probability:

[0130] ;

[0131] Calculate the change in behavior duration

[0132] Set the duration of normal behavior (seconds):

[0133] ;

[0134] Set the duration of abnormal behavior (seconds):

[0135] ;

[0136] Calculate the duration change at each time point:

[0137] ;

[0138] ;

[0139] ;

[0140] ;

[0141] ;

[0142] Calculate the average value of the behavior duration change:

[0143] ;

[0144] Calculate the behavior transition break rate

[0145] ;

[0146] Finally, calculate the behavior transition break rate , which indicates that there are significant changes in both the transition probability and duration of abnormal behavior compared to normal behavior. If it exceeds the set threshold (such as 2.0), it means that the abnormal behavior has a greater disturbing effect on the normal behavior pattern.

[0147] Based on the behavior transition break rate, the abnormal behavior impact assessment sub-module calculates the impact proportion of abnormal behavior on the normal behavior pattern, extracts the impact of abnormal behavior on normal behavior, analyzes the extended range of abnormal behavior on the time axis, analyzes the impact degree of abnormal behavior on behavior stability, establishes a measure index for the impact of the behavior pattern by abnormal behavior, and obtains the impact proportion of abnormal behavior;

[0148] Based on the behavior transition break rate, calculate the impact proportion of abnormal behavior on the normal behavior pattern, extract the impact of abnormal behavior on normal behavior, first calculate the extended range of abnormal behavior, and set the time step Record the propagation trend of abnormal behavior over multiple time steps. If the duration of abnormal behavior exceeds twice the average duration of normal behavior, calculate the influence weight of abnormal behavior on subsequent behavior, evaluate the influence proportion of abnormal behavior in the overall behavior pattern, establish an abnormal influence measure index for the behavior pattern, and compare the path similarity between normal behavior and abnormal behavior. If the path deviation of abnormal behavior exceeds the set threshold, record the influence weight of abnormal behavior, and finally obtain the influence proportion of abnormal behavior.

[0149] Please refer to Figure 6 , the behavior index update module includes:

[0150] The behavior interval analysis sub-module calls the influence proportion of abnormal behavior, obtains the duration sequence of the dog's behavior, calculates the behavior interval time based on the behavior data, calculates the change rate of the behavior interval, analyzes the correlation between the recurrence probability of abnormal behavior and the normal behavior interval, extracts the change trend of the behavior interval, and uses the formula:

[0151] ;

[0152] Calculate the change ratio of the behavior interval , and obtain the change trend of the behavior interval;

[0153] Among them, represents the change ratio of the behavior interval, is the interval time of the th behavior, is the interval time of the previous behavior, is the total number of behavior occurrences;

[0154] First, the system calls the influence proportion of abnormal behavior and extracts the duration sequence of the dog's behavior from historical data, including the time interval between each behavior occurrence. The data source can be a behavior sensor worn by the dog or manual observation records, and these data may include: movement behaviors (such as running, walking, standing), rest behaviors (such as lying down, stationary), and abnormal behaviors (such as irritability, excessive movement);

[0155] To ensure data integrity, the preprocessing stage includes: removing outliers: if some interval times deviate significantly from the overall data (such as more than 3 standard deviations above the mean), then remove that data point; time normalization: since different dogs have different activity levels, the time intervals need to be converted to relative values so that data from different individuals can be compared and analyzed; data sorting: sort by timestamp to ensure the consistency of the time series when calculating the change trend of the behavior interval.

[0156] Calculate the behavior interval time:

[0157] For a certain dog, its behavior interval data can be expressed as:

[0158] ;

[0159] Wherein, represents the time interval of the th behavior, represents the time interval of the previous behavior, is the total number of behaviors.

[0160] Calculate the change ratio of each behavior interval:

[0161] ;

[0162] For example, assume the behavior intervals of a certain dog are as follows: seconds, seconds, seconds, seconds

[0163] Then calculate the change ratio of each time interval:

[0164] ;

[0165] Calculate the change ratio of the behavior interval:

[0166] Use the given formula:

[0167] ;

[0168] Calculate the average change ratio of the behavior interval. For example, the calculation process of the above data:

[0169] ;

[0170] ;

[0171] The calculated indicates the degree of fluctuation of the time interval of the dog's behavior. If this value is large, it means that the behavior interval fluctuates greatly, and the dog may be in an unstable behavior pattern, such as being stimulated externally or having anxiety behaviors. If this value is small, it means that the dog's behavior interval is relatively stable and conforms to the normal behavior pattern.

[0172] The behavior matching degree evaluation sub-module constructs a behavior persistence weight matrix based on the change trend of the behavior interval, calculates the matching degree of abnormal behaviors in the overall behavior pattern, evaluates the matching degree, compares the recurrence probability of normal behaviors, screens candidate behavior data that conform to the behavior pattern, and obtains the behavior matching evaluation result;

[0173] The behavior matching degree evaluation sub-module begins to establish a behavior persistence weight matrix by analyzing the canine behavior data collected by the intelligent dog trainer, including the changing trend of its behavior intervals. The construction of this matrix is based on the occurrence frequency and duration of each behavior to accurately calculate the weight of each behavior. For example, if a dog often performs the sitting behavior during training and rarely shows the jumping behavior, the weight matrix will reflect this behavior pattern, with the weight of the sitting behavior higher than that of the jumping. Subsequently, the system uses this matrix to calculate the matching degree between the abnormal behavior and the normal behavior pattern, by comparing the statistical probability differences between the abnormal behavior and the common behavior to evaluate its matching degree. Finally, the system will screen out the candidate behavior data that matches the normal behavior pattern, which includes but is not limited to those behaviors that occur occasionally but are still within the normal range, thereby generating the behavior matching evaluation result. For example, for a dog that suddenly starts to frequently try to escape, the system may mark this behavior as high-risk and require further intervention.

[0174] The behavior index screening sub-module calls the behavior matching evaluation result, filters the data nodes of the behavior index, calculates the influence range of the behavior after the index update, compares the behavior change ratio before and after the update of the abnormal behavior, screens the behavior index update scheme, and obtains the intervention strategy after the behavior update.

[0175] Call the behavior matching degree evaluation result and filter the behavior data nodes. This process includes calculating the influence range of each data node to determine which behavior nodes need to be updated. For example, if a dog's escape behavior is marked as abnormal, the relevant behavior nodes will be identified as needing to be updated to modify the behavior pattern and training method. Subsequently, the system will calculate the influence range of the behavior after the index update and evaluate the behavior change ratio by comparing the data before and after the update of the abnormal behavior. This calculation helps to determine the most effective behavior intervention strategy, such as increasing specific training tasks or adjusting the training intensity. Finally, the selected behavior index update scheme will be used to optimize the training program and obtain the intervention strategy after the behavior update, which may include fine-tuning the training method or re-evaluating the dog's behavior pattern to ensure that the training effect is consistent with the expectation.

[0176] The above is only the preferred embodiment of the present invention and does not limit the present invention in other forms. Any person skilled in the art may use the disclosed technical content to make changes or modifications into equivalent embodiments with equivalent changes and apply them to other fields. However, as long as it does not depart from the technical solution content of the present invention, any simple modification, equivalent change, and modification made to the above embodiments based on the technical essence of the present invention still fall within the protection scope of the technical solution of the present invention.

Claims

1. An animal behavior management system based on an intelligent dog trainer, characterized in that, The system includes: The behavior data acquisition module obtains the behavior data of the dog, calls the sensor data of the dog trainer, records the movement trajectory and environmental factors, and screens the environmental behavior characteristic data; The high-dimensional offset calculation module extracts time series features based on the environmental behavior characteristic data, calculates the trajectory position difference and offset rate, records the current behavior path, judges the change rate of the offset, and combines the movement acceleration and direction change to screen the behavior transition rate, generating a dynamic offset prediction result; The behavior boundary recognition module calls the dynamic offset prediction result, records the continuity break of abnormal behaviors, divides time periods, calculates the movement frequency, sprint ratio, and stationary time, analyzes the change of movement rhythm, identifies the deviation of the behavior path, and generates a behavior fuzzy interval; The abnormal impact analysis module evaluates the impact of abnormal behaviors on the normal mode based on the behavior fuzzy interval, calculates the decrease in behavior transfer probability, behavior delay, and break rate, and generates the abnormal behavior impact ratio; The behavior index update module calls the abnormal behavior impact ratio, analyzes the change of behavior interval, calculates the new behavior matching degree, compares the probability of repeated behaviors, screens the index update data, and generates an intervention strategy after behavior update.

2. The animal behavior management system based on the intelligent dog trainer according to claim 1, characterized in that, The environmental behavior characteristic data includes movement characteristic data, trajectory data, and environmental characteristic data. The dynamic offset prediction result includes trajectory change characteristics, offset rate characteristics, and abnormal behavior offset values. The behavior fuzzy interval includes behavior continuity characteristics, time partition characteristics, and behavior path change characteristics. The abnormal behavior impact ratio includes behavior transfer characteristics, abnormal behavior impact characteristics, and behavior break characteristics. The intervention strategy after behavior update includes behavior interval characteristics, new behavior matching characteristics, and behavior index update characteristics.

3. The animal behavior management system based on the intelligent dog trainer according to claim 1, characterized in that, The behavior data acquisition module includes: The movement data extraction sub-module calls the built-in sensor data of the intelligent dog trainer, obtains the movement acceleration, behavior sequence, and activity frequency of the dog, calculates the behavior duration, statistically analyzes the rhythm change rate and the stability of adjacent behavior sequences, analyzes the movement trajectory points of the dog, records the number of sprints and the ratio of stationary time, and generates movement characteristic data; The behavior pattern analysis sub-module identifies the continuous behavior pattern of the dog based on the movement characteristic data, analyzes the behavior transition trend, extracts the occurrence probability of the behavior, calculates the correlation degree between different behaviors of the dog, and uses the formula: ； Performs calculations to obtain the stability coefficient of the dog's behavior sequence, and combines the behavior transition trend to obtain the behavior pattern characteristics; Among them, represents the behavior pattern feature, represents the duration of a single behavior, represents the behavior transition interval, is the total number of behaviors, is the time decay coefficient, represents the total duration of the behavior sequence, represents the sum of the products of the durations of all behaviors and the behavior transition intervals; The environmental feature screening sub-module calculates the impact of environmental feature changes on the behavior pattern based on the behavior pattern characteristics, combines the noise level, light intensity, and ground vibration conditions in the space where the dog is located, screens the environmental variables with high correlation, and obtains the environmental behavior characteristic data.

4. The animal behavior management system based on the intelligent dog trainer according to claim 1, wherein The high-dimensional offset calculation module includes: The trajectory position calculation sub-module extracts time series behavior characteristics based on the environmental behavior characteristic data, obtains the movement trajectory coordinates of adjacent time points, calculates the trajectory position change value within each time step, sets a time interval, calculates the change amount of the trajectory vector, and combines the trajectory data to obtain the trajectory position difference; The offset rate analysis sub-module calculates the offset rate between adjacent behavioral feature vectors based on the trajectory position difference, using the formula: ； Operate to obtain the behavioral offset rate, compare the behavioral trajectory data with the current behavioral path data, and obtain the change rate of the behavioral offset; Among them, represents the behavior offset rate, represents the trajectory coordinate value at the th time point, represents the target trajectory coordinate value at the th time point, is the total number of trajectory points within the time step, is the trajectory offset adjustment factor, represents the offset trend of the behavior trajectory, is the correction amount of the behavior vector change rate; It should be noted that there seems to be a duplicate label in the original text. You may want to check and correct it if necessary. The abnormal behavior screening sub-module calculates the abnormal behavior offset value based on the change rate of the behavioral offset, combines the dog's movement acceleration and direction change, screens the data with a behavior transition rate exceeding 0.8, and obtains the dynamic offset prediction result.

5. The animal behavior management system based on the intelligent dog trainer according to claim 1, characterized in that, The behavioral boundary recognition module includes: The behavior continuity calculation sub-module extracts the behavior sequences before and after the abnormal behavior based on the dynamic offset prediction result, calculates the time interval between consecutive behaviors, counts the behavior disruption rate, divides the behavior time period, and obtains the behavior continuity disruption value; The movement rhythm analysis sub-module calculates the movement frequency, sprint ratio, and stationary duration for each time period based on the behavior continuity disruption value, constructs a behavior change matrix within the time period using the multi-scale behavior distribution analysis method, analyzes the stability of the movement state in the time partition, using the formula: ； Operate to obtain the change amplitude of the movement rhythm, compare the calculation results of adjacent time periods, construct a behavior transfer trend vector, establish a time dynamic transfer matrix, and obtain the behavior transfer change coefficient; in, Represents the amplitude of the movement rhythm change, Representative The movement frequency value at each time point, Representative The duration of stillness at a time point, is the total number of behavior records within the time step, is the behavioral volatility weight, Representative The rate of change of behavior at a time point, represents the behavior change cycle parameter, is the behavior change smoothing factor, Represents the degree of fluctuation of the behavior range, is the motion adjustment factor, Represents the change parameter of sprint behavior proportion, A dispersion correction parameter representing the behavior rhythm within a time period; The behavioral path comparison sub-module compares the normal behavioral path with the abnormal behavioral path based on the behavior transfer change coefficient, calculates the path fitting degree index, screens the same type of path intervals, and obtains the behavioral fuzzy interval.

6. The animal behavior management system based on the intelligent dog trainer according to claim 1, wherein The abnormal impact analysis module includes: The behavior sequence stability evaluation sub-module extracts the behavior sequences before and after the abnormal behavior based on the behavioral fuzzy interval, compares the behavior stability before and after the abnormality, calculates the distribution probability of behavior categories within the time period, calculates the behavior frequency fluctuation, judges the disturbance degree of the behavior sequence, and calculates the behavior stability decline ratio according to the disturbance measurement standard to obtain the behavior sequence stability offset value; The behavior transfer impact calculation sub-module calculates the decrease amplitude of the behavior transfer probability within the abnormal behavior period based on the behavior sequence stability offset value, calculates the conversion rate from normal behavior to abnormal behavior, constructs a time-segmented behavior impact model, and calculates the dynamic impact of the abnormal behavior on the normal behavior sequence, using the formula: ； Operate to obtain the behavior transfer break rate, analyze the continuous interference impact of the abnormal behavior on the normal behavior in the behavior pattern, and calculate the overall change amplitude of the behavior pattern; Among them, represents the behavior transfer fracture rate, represents the normal behavior transfer probability at the th time point, represents the transfer probability when abnormal behavior occurs, is the total number of behavior records within the time step, represents the duration of normal behavior, represents the duration of abnormal behavior; The abnormal behavior impact evaluation sub-module calculates the impact ratio of the abnormal behavior on the normal behavior pattern based on the behavior transfer break rate, extracts the impact of the abnormal behavior on the normal behavior, analyzes the expansion range of the abnormal behavior on the time axis, analyzes the impact degree of the abnormal behavior on the behavior stability, and establishes a measure index for the behavior pattern affected by the abnormality to obtain the abnormal behavior impact ratio.

7. The animal behavior management system based on the intelligent dog trainer according to claim 1, characterized in that The behavioral index update module includes: The behavior interval analysis sub-module calls the abnormal behavior impact ratio, obtains the duration sequence of the dog's behavior, calculates the behavior interval time based on the behavior data, calculates the behavior interval change rate, analyzes the correlation between the recurrence probability of abnormal behavior and the normal behavior interval, extracts the behavior interval change trend, using the formula: ； Calculate the change ratio of behavior intervals to obtain the change trend of behavior intervals; Among them, represents the ratio of the change in the behavior interval, is the interval time of the th behavior, is the interval time of the previous behavior, is the total number of times the behavior occurs; Based on the changing trend of the behavior interval, the behavior matching degree evaluation sub-module constructs a behavior persistence weight matrix, calculates the matching degree of abnormal behaviors in the overall behavior pattern, evaluates the matching degree, compares the recurrence probability of normal behaviors, screens candidate behavior data that conform to the behavior pattern, and obtains the behavior matching evaluation result; The behavior index screening sub-module calls the behavior matching evaluation result, filters data nodes of the behavior index, calculates the behavior influence range after index update, compares the behavior change ratio before and after abnormal behavior update, screens the behavior index update scheme, and obtains the intervention strategy after behavior update.

Images