User behavior analysis method and system for offline stores based on temporal network

Abstract

The present invention relates to the field of data science and analysis, and discloses a method for analyzing user behavior of offline stores based on a time series network, including: analyzing store elements of offline stores, analyzing the behavioral relationship between store elements and store users corresponding to the offline stores, and determining potential user behaviors of store users; analyzing the behavior collection requirements of potential user behaviors, configuring a data sensor network of store users, calculating the behavior coverage coefficient and user interference coefficient of the data sensor network, optimizing the data sensor network, and obtaining a target data sensor network; using a video motion capture function to capture the video behavior features of store users, extracting the audio behavior features of store users, constructing a GRU time series network model of store users, and defining a time series attention mechanism for store users, and using the GRU time series network model and the time series attention mechanism to analyze the target user behaviors of store users. The present invention can improve the accuracy of offline store user behavior analysis.

Description

Technical Field

[0001] The present invention relates to a method and system for analyzing user behavior of offline stores based on a temporal network, and belongs to the field of data science and analysis. Background Art

[0002] User behavior analysis of offline stores refers to the process of observing, recording, interpreting and predicting customer behavior in physical stores through a series of methods and technical means. This analysis aims to gain a deep understanding of customer behavior patterns, preferences, needs and experiences so that stores can better meet customer needs, optimize operating strategies, and improve sales performance and customer satisfaction.

[0003] Traditional offline store user behavior analysis usually relies on manual observation, simple traffic counters and questionnaires to collect customer information. These methods often lack depth and accuracy and cannot fully capture the complex behavior patterns of customers, thus failing to provide stores with accurate operation optimization strategies and customer experience improvement plans. Summary of the invention

[0004] The present invention provides a method and system for realizing user behavior analysis of offline stores based on a temporal network, the main purpose of which is to improve the accuracy of user behavior analysis of offline stores.

[0005] To achieve the above purpose, the present invention provides a method for analyzing user behavior of offline stores based on a time series network, comprising:

[0006] Acquire store scene characteristics and historical store behavior data of offline stores, analyze store elements of the offline stores based on the store scene characteristics, analyze the behavior relationship between the store elements and the corresponding store users of the offline stores based on the historical store behavior data, and determine the potential user behavior of the store users based on the behavior relationship;

[0007] Analyze the behavior collection requirements of the potential user behaviors, configure the data sensor network of the store users based on the behavior collection requirements, calculate the behavior coverage coefficient and user interference coefficient of the data sensor network, optimize the data sensor network based on the behavior coverage coefficient, user interference coefficient and preset non-sensing data collection equipment, and obtain a target data sensor network;

[0008] Using the target data sensor network, collecting store behavior data of the offline stores, wherein the store behavior data includes video behavior data and audio behavior data;

[0009] Establishing a video motion capture function of the video behavior data, and based on the video behavior data, using the video motion capture function to capture the video behavior features of the store user, wherein the video behavior features include user posture, user running trajectory, and user emotion;

[0010] Based on the audio behavior data, the audio behavior features of the store users are extracted, wherein the audio behavior features include user audio content, user audio intonation, and user audio intensity; based on the historical store behavior data, a GRU timing network model of the store users is constructed, and a timing attention mechanism of the store users is defined; based on the video behavior features and the audio behavior features, the target user behavior of the store users is analyzed using the GRU timing network model and the timing attention mechanism.

[0011] Optionally, analyzing the behavior relationship between the store elements and the store users corresponding to the offline stores according to the historical store behavior data includes:

[0012] Obtaining element-related data of the store element;

[0013] Time-synchronize the element-related data and the historical store behavior data to obtain a synchronized data set;

[0014] Determining a user behavior type of the synchronized data set;

[0015] Based on the user behavior type and the synchronized data set, construct a behavior interaction graph of the store elements;

[0016] According to the behavior interaction graph, the behavior relationship between the store elements and the corresponding store users of the offline stores is analyzed.

[0017] Optionally, configuring the data sensor network of the store user based on the behavior collection requirement includes:

[0018] Based on the behavior collection requirements, marking the user interaction points of the store users;

[0019] Determine the behavior data collection device of the store user according to the behavior collection requirements and the user interaction points;

[0020] Establishing a device layout of the behavior data collection device at the user interaction point;

[0021] Identify a device network port of the behavior data collection device;

[0022] Based on the device network port and the device layout, construct a behavior network structure of the store user;

[0023] The data sensor network of the store users is configured through the behavioral network structure.

[0024] Optionally, the calculating the behavior coverage coefficient and the user interference coefficient of the data sensor network includes:

[0025] Identifying device parameters of the data sensor network corresponding to a user interaction point;

[0026] Based on the device parameters, analyzing the device coverage area of ​​the user interaction point;

[0027] defining a coverage weight of the user interaction point;

[0028] According to the coverage weight and the device coverage area, the behavior coverage coefficient of the data sensor network is calculated using the following formula:

[0029]

[0030] in, represents the behavioral coverage coefficient of the data sensor network, represents the number of user interaction points corresponding to the data sensor network, Represents the data sensor network The coverage weight of user interaction points, Represents the data sensor network The device coverage area of ​​each user interaction point, Represents the data sensor network The reference area of ​​each user interaction point;

[0031] defining a user interference factor of the data sensor network;

[0032] Determining a quantized interference factor value of the user interference factor;

[0033] Based on the interference factor quantized value of the user interference factor, a user interference coefficient of the data sensor network is calculated.

[0034] Optionally, the calculating the user interference coefficient of the data sensor network based on the interference factor quantization value of the user interference factor includes:

[0035] defining an interference factor weight of the user interference factor;

[0036] Analyze the interaction point interference weight of the user interaction point corresponding to the user interference factor;

[0037] Based on the interference factor weight, the interference factor quantization value and the interaction point interference weight, the user interference coefficient of the data sensor network is calculated using the following formula:

[0038]

[0039] in, User interference coefficient of data sensor network, represents the number of user interaction points corresponding to the data sensor network, Represents the data sensor network The interaction point interference weight of each user interaction point, Represents the data sensor network User interaction point The interference factor weight of each user’s interference factor, Represents the data sensor network User interaction point The interference factor quantization value of the user interference factor, Represents the data sensor network The number of user interference factors per user interaction point.

[0040] Optionally, based on the video behavior data, capturing the video behavior characteristics of the store user using the video motion capture function includes:

[0041] Extracting the subject of the video behavior data;

[0042] Constructing a three-dimensional heat map of the subject;

[0043] Based on the three-dimensional heat map, the key point coordinates of the store user's body are identified using the key point recognition sub-function corresponding to the video motion capture function, wherein the key point recognition sub-function:

[0044]

[0045] in, Represents key points of the human body The coordinates of the key points of the human body, Indicates search The largest value, Represents the horizontal coordinate value of the three-dimensional heat map. Represents the vertical coordinate value of the three-dimensional heat map. Represents the index of the corresponding row of the three-dimensional heat map, Indicates the index of the corresponding column of the three-dimensional heat map, Represents key points of the human body , Indicates the size of the domain, Represents a three-dimensional heat map;

[0046] Marking the key points of the store user's body according to the coordinates of the key points of the body;

[0047] Analyzing the user posture of the store user based on the human body key points;

[0048] Marking the key point coordinates of the continuous frames of the key points of the human body through the video behavior data;

[0049] Connecting the key point coordinates of the consecutive frames to obtain the user running trajectory of the store user;

[0050] Extracting a face image from the video behavior data;

[0051] By using the facial image, analyzing the user emotion of the store user using the emotion classification function corresponding to the video motion capture function;

[0052] Based on the user posture, the user running trajectory and the user emotion, the video behavior characteristics of the store user are determined.

[0053] Optionally, analyzing the user emotion of the store user by using the facial image and the emotion classification function corresponding to the video motion capture function includes:

[0054] De-noising the face image to obtain a de-noised face image;

[0055] Correcting the denoised face image to obtain a corrected face image;

[0056] Analyzing the feature vector of the corrected face image;

[0057] Defining a user emotion type of the store user;

[0058] Based on the feature vector and the user emotion type, the emotion classification function is used to analyze the user emotion probability of the store user:

[0059]

[0060] in, Represented in the feature vector User emotion type The probability of user emotions, Represents the feature vector , Indicates the user's emotion type , represents the emotion classification function, represents the weight matrix, represents the bias term, represents the transpose of the weight matrix;

[0061] The user emotion of the store user is determined according to the user emotion probability.

[0062] Optionally, extracting the audio behavior features of the store user based on the audio behavior data includes:

[0063] Remove background noise from the audio behavior data to obtain a denoised audio signal;

[0064] Converting the de-noised audio signal into a signal text;

[0065] Extracting keywords from the signal text, and analyzing user audio content of the store user through the keywords;

[0066] Segmenting the denoised audio signal into individual phonemes;

[0067] analyzing frequency changes of the single phoneme;

[0068] Analyzing the user audio intonation of the store user according to the frequency change;

[0069] Calculating the audio energy of the denoised audio signal;

[0070] Based on the audio energy, determining a user audio intensity of the store user;

[0071] The audio behavior characteristics of the store user are determined based on the user audio content, the user audio tone, and the user audio intensity.

[0072] Optionally, constructing a GRU temporal network model of the store users based on the historical store behavior data includes:

[0073] Dividing the historical store behavior data into a training set and a validation set;

[0074] Define the GRU temporal network model architecture of the store user, wherein the GRU temporal network model architecture includes an input layer, a GRU layer, an output layer, and an activation function;

[0075] Using the training set, training the GRU temporal network model architecture to obtain a trained GRU temporal network model;

[0076] Using the validation set, the performance of the trained GRU temporal network model is verified to obtain the model performance;

[0077] When the model performance meets the preset model performance threshold, the trained GRU timing network model is used as the GRU timing network model of the store user.

[0078] In order to solve the above problems, the present invention also provides a user behavior analysis system for offline stores based on a time series network, the system comprising:

[0079] A potential user behavior analysis module is used to obtain store scene characteristics and historical store behavior data of offline stores, analyze store elements of the offline stores based on the store scene characteristics, analyze the behavior relationship between the store elements and the corresponding store users of the offline stores based on the historical store behavior data, and determine the potential user behavior of the store users based on the behavior relationship;

[0080] A data sensor network construction module is used to analyze the behavior collection requirements of the potential user behaviors, configure the data sensor network of the store users based on the behavior collection requirements, calculate the behavior coverage coefficient and user interference coefficient of the data sensor network, optimize the data sensor network based on the behavior coverage coefficient, user interference coefficient and preset non-sensing data collection equipment, and obtain a target data sensor network;

[0081] A behavior data collection module, used to collect store behavior data of the offline stores using the target data sensor network, wherein the store behavior data includes video behavior data and audio behavior data;

[0082] A video behavior feature analysis module, used to establish a video motion capture function of the video behavior data, and based on the video behavior data, use the video motion capture function to capture the video behavior features of the store user, wherein the video behavior features include user posture, user running trajectory, and user emotion;

[0083] A user behavior analysis module is used to extract the audio behavior features of the store users based on the audio behavior data, wherein the audio behavior features include user audio content, user audio intonation and user audio intensity; based on the historical store behavior data, a GRU timing network model of the store users is constructed, and a timing attention mechanism of the store users is defined; based on the video behavior features and the audio behavior features, the target user behavior of the store users is analyzed using the GRU timing network model and the timing attention mechanism.

[0084] In order to solve the above problem, the present invention further provides an electronic device, the electronic device comprising:

[0085] at least one processor; and,

[0086] a memory communicatively connected to the at least one processor; wherein,

[0087] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to implement the above-mentioned method for analyzing user behavior of offline stores based on a time series network.

[0088] In order to solve the above problems, the present invention also provides a computer-readable storage medium, in which at least one instruction is stored. The at least one instruction is executed by a processor in an electronic device to implement the above-mentioned method for analyzing user behavior of offline stores based on a time series network.

[0089] Compared with the problems described in the background technology, the present invention analyzes the behavioral relationship between the store elements and the corresponding store users of the offline stores based on the historical store behavior data, so as to better understand how the store elements affect user behavior; the present invention determines the potential user behavior of the store users based on the behavioral relationship, so as to clarify the actions that the store users will take, thereby improving the completeness of action collection; further, the present invention can better arrange the behavior collection equipment and improve the reliability of data collection by analyzing the behavior collection requirements of the potential user behavior. Furthermore, the present invention can quantify the behavior coverage degree and user interference degree of the sensor network by calculating the behavior coverage coefficient and user interference coefficient of the data sensor network, thereby providing a basis for network optimization. Furthermore, the present invention optimizes the data sensor network based on the behavior coverage coefficient, user interference coefficient and preset non-sensing data collection equipment to obtain the target data sensor network, which can ensure that its data sensor network will not have a negative impact on the customer's shopping experience while collecting valuable data. Furthermore, the present invention uses the target data sensor network to collect the offline store user's behavior. The store behavior data of the offline stores can more comprehensively understand the behavior and needs of customers, thereby improving the operating efficiency of the stores. Next, the video action capture function of the video behavior data established by the present invention can realize the behavior feature capture under the video data of the store users. The present invention extracts the audio behavior features of the store users based on the audio behavior data, wherein the audio behavior features include user audio content, user audio intonation and user audio intensity, which can extract feature content from the audio. Combined with the above video features, the accuracy of behavior analysis is improved. Finally, the present invention builds the GRU time series network model of the store users based on the historical store behavior data, which can process the time dynamics in the time series data, better understand the current behavior state and define the time series attention mechanism of the store users. Based on the video behavior features and the audio behavior features, the target user behavior of the store users can be analyzed by using the GRU time series network model and the time series attention mechanism, which can help the model focus more on the most important time step for predicting the target user behavior, thereby improving the prediction accuracy and effect of the model for user behavior analysis. Therefore, the present invention can improve the accuracy of offline store user behavior analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0090] Figure 1A flow chart of a method for analyzing user behavior of offline stores based on a time series network provided in one embodiment of the present invention;

[0091] Figure 2 A functional module diagram of a user behavior analysis system for offline stores based on a time series network provided in one embodiment of the present invention;

[0092] Figure 3 A schematic diagram of the structure of an electronic device for implementing a user behavior analysis system for offline stores based on a time series network provided by an embodiment of the present invention;

[0093] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings in conjunction with the embodiments. DETAILED DESCRIPTION

[0094] 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.

[0095] The embodiment of the present application provides a method for analyzing user behavior of offline stores based on a timing network. The execution subject of the method for analyzing user behavior of offline stores based on a timing network includes but is not limited to at least one of the electronic devices such as a server and a terminal that can be configured to execute the method provided by the embodiment of the present application. In other words, the method for analyzing user behavior of offline stores based on a timing network can be executed by software or hardware installed on a terminal device or a server device. The server includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, etc.

[0096] Embodiment 1:

[0097] Reference Figure 1 FIG. 1 is a flow chart of a method for analyzing user behavior of offline stores based on a time series network according to an embodiment of the present invention. In this embodiment, the method for analyzing user behavior of offline stores based on a time series network includes:

[0098] S1. Acquire store scene characteristics and historical store behavior data of offline stores, analyze store elements of the offline stores based on the store scene characteristics, analyze the behavioral relationship between the store elements and the corresponding store users of the offline stores based on the historical store behavior data, and determine the potential user behavior of the store users based on the behavioral relationship.

[0099] It should be explained that the offline stores refer to physical retail stores, including but not limited to supermarkets, specialty stores, department stores, restaurants, etc. The store scene features refer to the specific environment and layout characteristics of offline stores, including spatial layout, product display, customer flow, etc. The historical store behavior data refers to the customer behavior activity data recorded and accumulated in the past period of time, usually including customer flow data, customer behavior data, transaction data, interaction data, etc. The store elements refer to the various physical and psychological elements that constitute the store environment and customer experience.

[0100] The present invention analyzes the behavioral relationship between the store elements and the store users corresponding to the offline stores based on the historical store behavior data, so as to better understand how the store elements affect the user behavior.

[0101] In detail, analyzing the behavior relationship between the store elements and the store users corresponding to the offline stores according to the historical store behavior data includes:

[0102] Obtaining element-related data of the store element;

[0103] Time-synchronize the element-related data and the historical store behavior data to obtain a synchronized data set;

[0104] Determining a user behavior type of the synchronized data set;

[0105] Based on the user behavior type and the synchronized data set, construct a behavior interaction graph of the store elements;

[0106] According to the behavior interaction graph, the behavior relationship between the store elements and the corresponding store users of the offline stores is analyzed.

[0107] Among them, the element-related data refers to data directly related to store elements, including but not limited to the store's physical layout, product display, lighting conditions, audio settings, logo design and other data; the synchronized data set refers to a data set formed by aligning the element-related data with the historical store behavior data in time; the user behavior type refers to customer behavior that can be observed and classified in the store, such as browsing products, walking, talking and other types; the behavior interaction map shows a graphical representation of the interaction and relationship between store elements and user behavior types; the behavior relationship refers to the description of the correlation or causal relationship between store elements and user behavior.

[0108] Optionally, the construction of the behavior interaction graph of the store elements based on the user behavior type and the synchronized data set may be constructed using graph theory and network analysis techniques.

[0109] Based on the behavioral relationship, the present invention determines the potential user behavior of the store user to clarify the actions that the store user will perform, thereby improving the completeness of action collection, wherein the potential user behavior refers to user behavior that may occur, such as browsing products, consulting, trying on, purchasing, returning products, staying in a specific area, etc.

[0110] S2. Analyze the behavior collection requirements of the potential user behavior, configure the data sensor network of the store users based on the behavior collection requirements, calculate the behavior coverage coefficient and user interference coefficient of the data sensor network, optimize the data sensor network based on the behavior coverage coefficient, user interference coefficient and the preset non-contact data collection equipment, and obtain the target data sensor network.

[0111] The present invention can better arrange behavior collection equipment and improve the reliability of data collection by analyzing the behavior collection requirements of the potential user behavior. The behavior collection requirements refer to the requirements for collecting the behavior of the potential user, such as video collection equipment is required for collecting user actions, audio collection equipment is required for collecting user voices, behavior data collection locations and other requirements.

[0112] Based on the behavior collection requirement, the present invention configures the store user's data sensor network to achieve efficient collection of store user's actions.

[0113] In detail, configuring the data sensor network of the store users based on the behavior collection requirements includes:

[0114] Based on the behavior collection requirements, marking the user interaction points of the store users;

[0115] Determine the behavior data collection device of the store user according to the behavior collection requirements and the user interaction points;

[0116] Establishing a device layout of the behavior data collection device at the user interaction point;

[0117] Identify a device network port of the behavior data collection device;

[0118] Based on the device network port and the device layout, construct a behavior network structure of the store user;

[0119] The data sensor network of the store users is configured through the behavioral network structure.

[0120] Among them, the user interaction point refers to the specific location or scene where customers interact in the store, such as the cash register, product display area, fitting room, information desk, etc. The behavior data collection device refers to the hardware device used to capture and record user behavior, such as cameras, Wi-Fi probes, RFID readers, QR code scanners, touch screens, etc. The equipment layout refers to the specific placement and distribution of the behavior data collection equipment in the store. The device network port refers to the interface on the behavior data collection device for connecting to the network and transmitting data, such as Ethernet port, Wi-Fi antenna, Bluetooth module, etc. The behavior network structure refers to the network architecture composed of multiple behavior data collection devices, including the connection method of the equipment, data flow direction, network topology, etc. The data sensor network refers to the overall system composed of behavior data collection equipment, transmission media, data processing center and other related components.

[0121] The present invention can quantify the behavior coverage degree and user interference degree of the sensor network by calculating the behavior coverage coefficient and user interference coefficient of the data sensor network, thereby providing a basis for network optimization.

[0122] In detail, the calculating of the behavior coverage coefficient and the user interference coefficient of the data sensor network includes:

[0123] Identifying device parameters of the data sensor network corresponding to a user interaction point;

[0124] Based on the device parameters, analyzing the device coverage area of ​​the user interaction point;

[0125] defining a coverage weight of the user interaction point;

[0126] According to the coverage weight and the device coverage area, the behavior coverage coefficient of the data sensor network is calculated using the following formula:

[0127]

[0128] in, represents the behavioral coverage coefficient of the data sensor network, represents the number of user interaction points corresponding to the data sensor network, Represents the data sensor network The coverage weight of user interaction points, Represents the data sensor network The device coverage area of ​​each user interaction point, Represents the data sensor network The reference area of ​​each user interaction point;

[0129] defining a user interference factor of the data sensor network;

[0130] Determining a quantized interference factor value of the user interference factor;

[0131] Based on the interference factor quantized value of the user interference factor, a user interference coefficient of the data sensor network is calculated.

[0132] The device parameters refer to the technical specifications of each device in the data sensor network, including but not limited to the type, sensing range, resolution and other parameters of the sensor. The device coverage area refers to the area where a single sensor can effectively detect or collect data. The coverage weight refers to the importance or priority assigned to each user interaction point, reflecting the relative importance of the interaction point in data collection. The behavior coverage coefficient refers to an indicator that measures the effectiveness of the data sensor network in covering user interaction points. The reference area refers to the ideal coverage area of ​​each user interaction point set for comparison. The user interference factor refers to negative factors that may affect the customer experience, such as sensor intrusion on customers, privacy concerns, device appearance, lighting, etc. For example, the camera flash or the red light is always on, which may cause visual interference caused by discomfort to customers, and auditory interference caused by beeps or other prompt sounds emitted by the sensor. The interference factor quantification value refers to the numerical representation of the specific impact degree of each user interference factor. The interference factor quantification value can be determined by a preset user interference factor scoring table. The user interference coefficient refers to an indicator that measures the degree of interference of the data sensor network on the customer experience.

[0133] Further, the calculating of the user interference coefficient of the data sensor network based on the interference factor quantization value of the user interference factor includes:

[0134] defining an interference factor weight of the user interference factor;

[0135] Analyze the interaction point interference weight of the user interaction point corresponding to the user interference factor;

[0136] Based on the interference factor weight, the interference factor quantization value and the interaction point interference weight, the user interference coefficient of the data sensor network is calculated using the following formula:

[0137]

[0138] in, User interference coefficient of data sensor network, represents the number of user interaction points corresponding to the data sensor network, Represents the data sensor network The interaction point interference weight of each user interaction point, Represents the data sensor network User interaction point The interference factor weight of each user’s interference factor, Represents the data sensor network User interaction point The interference factor quantization value of the user interference factor, Represents the data sensor network The number of user interference factors per user interaction point.

[0139] The interference factor weight refers to the relative importance of the interference factor assigned to each user, and the interaction point interference weight refers to the importance of the interference impact assigned to each user interaction point.

[0140] The present invention optimizes the data sensor network based on the behavior coverage coefficient, user interference coefficient and preset non-sensing data collection equipment, and obtains a target data sensor network that can ensure that its data sensor network will not have a negative impact on the customer's shopping experience while collecting valuable data. Among them, the non-sensing data collection equipment refers to those devices that have the least impact on user behavior and experience when collecting data, such as wireless cameras, infrared sensors and other devices, and the target data sensor network refers to a data sensor network that has been optimized to collect the required data in the most efficient and least disturbing way.

[0141] S3. Utilizing the target data sensor network, collecting store behavior data of the offline stores, wherein the store behavior data includes video behavior data and audio behavior data.

[0142] The present invention utilizes the target data sensor network to collect the store behavior data of the offline stores, so as to more comprehensively understand the behavior and needs of customers, thereby improving the operation efficiency of the stores.

[0143] It should be explained that the video behavior data refers to the customer's behavior and activity records in the store collected by the video surveillance system. These data usually include visual information such as the customer's movement path, stay time, interaction, facial expressions, etc. The audio behavior data refers to the sound information in the store collected by the audio sensor, including customer conversations, the voice of the store staff, background music, noise level and other information.

[0144] S4. Establish a video motion capture function of the video behavior data, and based on the video behavior data, use the video motion capture function to capture the video behavior characteristics of the store user, wherein the video behavior characteristics include user posture, user movement trajectory, and user emotion.

[0145] The video action capture function of the video behavior data established by the present invention can realize the behavioral feature capture under the video data of the store user. Among them, the video action capture function refers to a set of algorithms or programs for identifying and extracting action information from video data. It can be constructed based on the deep learning model of computer vision technology.

[0146] Based on the video behavior data, the present invention utilizes the video motion capture function to capture the video behavior characteristics of the store users and improves the accuracy of subsequent behavior determination through multi-angle video behavior analysis.

[0147] In detail, based on the video behavior data, the video action capture function is used to capture the video behavior characteristics of the store user, including:

[0148] Extracting the subject of the video behavior data;

[0149] Constructing a three-dimensional heat map of the subject;

[0150] Based on the three-dimensional heat map, the key point coordinates of the store user's body are identified using the key point recognition sub-function corresponding to the video motion capture function, wherein the key point recognition sub-function:

[0151]

[0152] in, Represents key points of the human body The coordinates of the key points of the human body, Indicates search The largest value, Represents the horizontal coordinate value of the three-dimensional heat map. Represents the vertical coordinate value of the three-dimensional heat map. Represents the index of the corresponding row of the three-dimensional heat map, Indicates the index of the corresponding column of the three-dimensional heat map, Represents key points of the human body , Indicates the size of the domain, Represents a three-dimensional heat map;

[0153] Marking the key points of the store user's body according to the coordinates of the key points of the body;

[0154] Analyzing the user posture of the store user based on the human body key points;

[0155] Marking the key point coordinates of the continuous frames of the key points of the human body through the video behavior data;

[0156] Connecting the key point coordinates of the consecutive frames to obtain the user running trajectory of the store user;

[0157] Extracting a face image from the video behavior data;

[0158] By using the facial image, analyzing the user emotion of the store user using the emotion classification function corresponding to the video motion capture function;

[0159] Based on the user posture, the user running trajectory and the user emotion, the video behavior characteristics of the store user are determined.

[0160] Among them, the character subject refers to the main character in the video, the three-dimensional heat map refers to a map used to display the distribution of data in three-dimensional space, the key point identification subfunction refers to the coordinates used to identify and extract the key points of the human body, such as the head, shoulders, elbows, wrists and other key points, the human body key point coordinates refer to the specific positions of the human body key points in the video frame, usually expressed by (x, y) coordinates, the human body key points refer to representative points on the human body, such as the head, shoulders, elbows, wrists, hips, knees and ankles, the user posture refers to the body posture of the user in the video, the continuous frame key point coordinates refer to the coordinate sequence of the human body key points in the continuous video frames, the user running trajectory refers to the movement path of the user in the video obtained by connecting the key point coordinates in the continuous frames, the emotion classification function refers to the function used to analyze the facial expressions of the characters in the video, the user emotion refers to the emotional state of the user in the video, such as happiness, sadness, anger and other emotions, and the video behavior feature refers to the behavior pattern of the user in the video obtained by analyzing the user's posture, running trajectory and emotion.

[0161] Further, the analyzing the user emotions of the store user by using the facial image and the emotion classification function corresponding to the video motion capture function includes:

[0162] De-noising the face image to obtain a de-noised face image;

[0163] Correcting the denoised face image to obtain a corrected face image;

[0164] Analyzing the feature vector of the corrected face image;

[0165] Defining a user emotion type of the store user;

[0166] Based on the feature vector and the user emotion type, the emotion classification function is used to analyze the user emotion probability of the store user:

[0167]

[0168] in, Represented in the feature vector User emotion type The probability of user emotions, Represents the feature vector , Indicates the user's emotion type , represents the emotion classification function, represents the weight matrix, represents the bias term, represents the transpose of the weight matrix;

[0169] The user emotion of the store user is determined according to the user emotion probability.

[0170] Among them, the denoised face image refers to an image with reduced noise and interference after the original face image is processed by applying a denoising algorithm (such as wavelet transform, non-local mean filtering, deep learning denoising network, etc.); the corrected face image refers to an image in a standard view by performing a geometric transformation (such as affine transformation, perspective transformation, etc.) on the denoised image to correct image distortion caused by factors such as camera angle and face posture; the feature vector refers to a set of values ​​extracted from the corrected face image; the user emotion type refers to a predefined emotion category, such as happiness, sadness, surprise, anger, fear, disgust and indifference; the user emotion probability refers to the probability that the user shows a certain specific emotion type under a given feature vector, which is usually a value between 0 and 1; the weight matrix refers to a parameter matrix in the emotion classification function used to map the feature vector to the emotion category; the bias term refers to another parameter in the emotion classification function, which is used to adjust the output probability; and the transpose of the weight matrix refers to transposing all elements of the weight matrix.

[0171] S5. Based on the audio behavior data, extract the audio behavior features of the store user, wherein the audio behavior features include user audio content, user audio intonation and user audio intensity; based on the historical store behavior data, construct a GRU timing network model of the store user, and define a timing attention mechanism for the store user; based on the video behavior features and the audio behavior features, use the GRU timing network model and the timing attention mechanism to analyze the target user behavior of the store user.

[0172] The present invention extracts the audio behavior features of the store users based on the audio behavior data, wherein the audio behavior features include user audio content, user audio intonation and user audio intensity. Feature content can be extracted from the audio and combined with the above-mentioned video features to improve the accuracy of behavior analysis.

[0173] In detail, extracting the audio behavior features of the store user based on the audio behavior data includes:

[0174] Remove background noise from the audio behavior data to obtain a denoised audio signal;

[0175] Converting the de-noised audio signal into a signal text;

[0176] Extracting keywords from the signal text, and analyzing user audio content of the store user through the keywords;

[0177] Segmenting the denoised audio signal into individual phonemes;

[0178] analyzing frequency changes of the single phoneme;

[0179] Analyzing the user audio intonation of the store user according to the frequency change;

[0180] Calculating the audio energy of the denoised audio signal;

[0181] Based on the audio energy, determining a user audio intensity of the store user;

[0182] The audio behavior characteristics of the store user are determined based on the user audio content, the user audio tone, and the user audio intensity.

[0183] Among them, the denoised audio signal refers to an audio signal with reduced background noise interference through noise reduction processing, the signal text refers to the text form of the denoised audio signal converted by automatic speech recognition (ASR) technology, the keyword refers to the representative and distinctive vocabulary extracted from the signal text, the user audio content refers to the overview of what the user said obtained by analyzing the keywords in the signal text and understanding the context, the single phoneme refers to the most basic speech unit segmented from the audio signal, the frequency change refers to the change of the pitch of a single phoneme over time, the user audio intonation refers to the pitch pattern of the user voice identified by analyzing the frequency change of the phoneme, the audio energy refers to the intensity or volume level of the audio signal, the user audio intensity refers to the loudness of the user voice calculated based on the audio energy, and the audio behavior feature refers to a series of features extracted from the audio data, including user audio content, user audio intonation and user audio intensity.

[0184] Optionally, analyzing the user audio content of the store users through the keywords may be achieved through natural language processing (NLP) technology.

[0185] Optionally, the determining of the user audio intensity of the store user based on the audio energy may be performed by analyzing audio intensity changes through a time window, such as short-time energy (STE) or short-time average amplitude (STA) analysis.

[0186] The present invention constructs a GRU time series network model of the store users based on the historical store behavior data, which can process the time dynamics in the time series data and better understand the current behavior status.

[0187] In detail, the GRU temporal network model of the store users is constructed based on the historical store behavior data, including:

[0188] Dividing the historical store behavior data into a training set and a validation set;

[0189] Define the GRU temporal network model architecture of the store user, wherein the GRU temporal network model architecture includes an input layer, a GRU layer, an output layer, and an activation function;

[0190] Using the training set, training the GRU temporal network model architecture to obtain a trained GRU temporal network model;

[0191] Using the validation set, the performance of the trained GRU temporal network model is verified to obtain the model performance;

[0192] When the model performance meets the preset model performance threshold, the trained GRU timing network model is used as the GRU timing network model of the store user.

[0193] Among them, the training set refers to a subset of historical store behavior data, which is used to train the model; the validation set refers to another part of historical store behavior data, which is used to evaluate the performance of the model during the model training process; the input layer refers to the first layer of the GRU time series network model, which is used to receive the input of time series data; the GRU layer refers to the core of the GRU time series network model, which contains multiple GRU units, each of which has a gating mechanism to control the flow of information; the output layer refers to the last layer of the GRU time series network model, which produces the final prediction or classification result; the activation function refers to the function applied to the GRU layer and output layer nodes, which is used to introduce nonlinear factors so that the model can learn and simulate more complex data relationships; the trained GRU time series network model refers to a model that minimizes the loss function of the model on the training set by iteratively adjusting the model parameters; the model performance refers to the accuracy of the model's prediction results on the validation set or test set; the model performance threshold refers to a pre-set performance standard that the model must meet or exceed to be considered effective; and the GRU time series network model refers to a deep learning model specifically used to process time series data.

[0194] Optionally, the GRU temporal network model architecture is trained using the training set, and the trained GRU temporal network model can be implemented through forward propagation and back propagation operations.

[0195] Finally, the present invention defines the temporal attention mechanism of the store user. Based on the video behavior features and the audio behavior features, using the GRU temporal network model and the temporal attention mechanism to analyze the target user behavior of the store user can help the model focus more on the most important time step for predicting the target user behavior, thereby improving the prediction accuracy and effect of the model's user behavior analysis. Among them, the temporal attention mechanism refers to a mechanism for weighting different time steps when processing sequence data, which enables the model to pay more attention to the part of the sequence that is most relevant to the current task. The target user behavior refers to the specific behavior of the store user during shopping, browsing, interaction, etc., such as purchasing goods, browsing specific areas, communicating with store clerks, etc.

[0196] Compared with the problems described in the background technology, the present invention analyzes the behavioral relationship between the store elements and the corresponding store users of the offline stores based on the historical store behavior data, so as to better understand how the store elements affect user behavior; the present invention determines the potential user behavior of the store users based on the behavioral relationship, so as to clarify the actions that the store users will take, thereby improving the completeness of action collection; further, the present invention can better arrange the behavior collection equipment and improve the reliability of data collection by analyzing the behavior collection requirements of the potential user behavior. Furthermore, the present invention can quantify the behavior coverage degree and user interference degree of the sensor network by calculating the behavior coverage coefficient and user interference coefficient of the data sensor network, thereby providing a basis for network optimization. Furthermore, the present invention optimizes the data sensor network based on the behavior coverage coefficient, user interference coefficient and preset non-sensing data collection equipment to obtain the target data sensor network, which can ensure that its data sensor network will not have a negative impact on the customer's shopping experience while collecting valuable data. Furthermore, the present invention uses the target data sensor network to collect the offline store user's behavior. The store behavior data of the offline stores can more comprehensively understand the behavior and needs of customers, thereby improving the operating efficiency of the stores. Next, the video action capture function of the video behavior data established by the present invention can realize the behavior feature capture under the video data of the store users. The present invention extracts the audio behavior features of the store users based on the audio behavior data, wherein the audio behavior features include user audio content, user audio intonation and user audio intensity, which can extract feature content from the audio. Combined with the above video features, the accuracy of behavior analysis is improved. Finally, the present invention builds the GRU time series network model of the store users based on the historical store behavior data, which can process the time dynamics in the time series data, better understand the current behavior state and define the time series attention mechanism of the store users. Based on the video behavior features and the audio behavior features, the target user behavior of the store users can be analyzed by using the GRU time series network model and the time series attention mechanism, which can help the model focus more on the most important time step for predicting the target user behavior, thereby improving the prediction accuracy and effect of the model for user behavior analysis. Therefore, the present invention can improve the accuracy of offline store user behavior analysis.

[0197] Embodiment 2:

[0198] like Figure 2 , which is a functional module diagram of a user behavior analysis system for offline stores based on a time series network provided by an embodiment of the present invention.

[0199] The user behavior analysis system 200 for offline stores based on a time series network of the present invention can be installed in an electronic device. According to the functions implemented, the user behavior analysis system 200 for offline stores based on a time series network can include a potential user behavior analysis module 201, a data sensor network construction module 202, a behavior data acquisition module 203, a video behavior feature analysis module 204 and a user behavior analysis module 205. The module of the present invention can also be referred to as a unit, which refers to a series of computer program segments that can be executed by an electronic device processor and can complete fixed functions, which are stored in the memory of the electronic device.

[0200] In this embodiment, the functions of each module / unit are as follows:

[0201] The potential user behavior analysis module 201 is used to obtain store scene characteristics and historical store behavior data of offline stores, analyze store elements of the offline stores based on the store scene characteristics, analyze the behavior relationship between the store elements and the store users corresponding to the offline stores based on the historical store behavior data, and determine the potential user behavior of the store users based on the behavior relationship;

[0202] The data sensor network construction module 202 is used to analyze the behavior collection requirements of the potential user behaviors, configure the data sensor network of the store users based on the behavior collection requirements, calculate the behavior coverage coefficient and user interference coefficient of the data sensor network, and optimize the data sensor network based on the behavior coverage coefficient, user interference coefficient and preset non-sensing data collection equipment to obtain a target data sensor network;

[0203] The behavior data collection module 203 is used to collect store behavior data of the offline stores using the target data sensor network, wherein the store behavior data includes video behavior data and audio behavior data;

[0204] The video behavior feature analysis module 204 is used to establish a video motion capture function of the video behavior data, and based on the video behavior data, use the video motion capture function to capture the video behavior features of the store user, wherein the video behavior features include user posture, user running trajectory, and user emotion;

[0205] The user behavior analysis module 205 is used to extract the audio behavior features of the store users based on the audio behavior data, wherein the audio behavior features include user audio content, user audio intonation and user audio intensity, construct a GRU timing network model of the store users based on the historical store behavior data, and define a timing attention mechanism for the store users, and analyze the target user behavior of the store users based on the video behavior features and the audio behavior features using the GRU timing network model and the timing attention mechanism.

[0206] In detail, the modules described in the user behavior analysis system 200 for offline stores based on a timing network described in the embodiment of the present invention adopt the same technical means as the user behavior analysis method for offline stores based on a timing network described in the accompanying drawings when used, and can produce the same technical effects, which will not be repeated here.

[0207] An embodiment of the present invention provides an electronic device for implementing a method for analyzing user behavior of an offline store based on a temporal network.

[0208] See also Figure 3 As shown, the electronic device may include a processor 30, a memory 31, a communication bus 32 and a communication interface 33, and may also include a computer program stored in the memory 31 and executable on the processor 30, such as a user behavior analysis method program for offline stores based on a timing network.

[0209] Among them, the processor can be composed of an integrated circuit in some embodiments, for example, it can be composed of a single packaged integrated circuit, or it can be composed of multiple integrated circuits with the same function or different functions, including one or more central processing units (CPU), microprocessors, digital processing chips, graphics processors and various control chips. The processor is the control core (Control Unit) of the electronic device, which uses various interfaces and lines to connect the various components of the entire electronic device, and executes or executes the program or module stored in the memory (for example, executing a user behavior analysis program for offline stores based on a timing network, etc.), and calls the data stored in the memory to execute various functions of the electronic device and process data.

[0210] The memory includes at least one type of readable storage medium, and the readable storage medium includes flash memory, mobile hard disk, multimedia card, card-type memory (for example: SD or DX memory, etc.), magnetic memory, disk, optical disk, etc. In some embodiments, the memory may be an internal storage unit of an electronic device, such as a mobile hard disk of the electronic device. In other embodiments, the memory may also be an external storage device of an electronic device, such as a plug-in mobile hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the electronic device. Further, the memory may also include both an internal storage unit and an external storage device of the electronic device. The memory can be used not only to store application software and various types of data installed in the electronic device, such as the code of the user behavior analysis program for offline stores based on the timing network, but also to temporarily store data that has been output or is to be output.

[0211] The communication bus may be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. The bus is configured to realize connection and communication between the memory and at least one processor, etc.

[0212] The communication interface is used for communication between the above-mentioned electronic device and other devices, including a network interface and a user interface. Optionally, the network interface may include a wired interface and / or a wireless interface (such as a WI-FI interface, a Bluetooth interface, etc.), which is generally used to establish a communication connection between the electronic device and other electronic devices. The user interface may be a display (Display), an input unit (such as a keyboard (Keyboard)), and optionally, the user interface may also be a standard wired interface, a wireless interface. Optionally, in some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, and an OLED (Organic Light-Emitting Diode, organic light-emitting diode) touch device, etc. Among them, the display may also be appropriately referred to as a display screen or a display unit, which is used to display information processed in the electronic device and to display a visual user interface.

[0213] For example, although not shown, the electronic device may also include a power source (such as a battery) for supplying power to various components. Preferably, the power source may be logically connected to the at least one processor through a power management system, so that the power management system can realize functions such as charging management, discharging management, and power consumption management. The power source may also include any components such as one or more DC or AC power sources, recharging systems, power failure detection circuits, power converters or inverters, and power status indicators. The electronic device may also include a variety of sensors, Bluetooth modules, Wi-Fi modules, etc., which will not be repeated here.

[0214] It should be understood that the embodiment is for illustration only and the scope of the patent application is not limited to this structure.

[0215] The user behavior analysis program for offline stores based on a time-series network stored in the memory of the electronic device is a combination of multiple instructions. When running in the processor, the following can be achieved:

[0216] Acquire store scene characteristics and historical store behavior data of offline stores, analyze store elements of the offline stores based on the store scene characteristics, analyze the behavior relationship between the store elements and the corresponding store users of the offline stores based on the historical store behavior data, and determine the potential user behavior of the store users based on the behavior relationship;

[0217] Analyze the behavior collection requirements of the potential user behaviors, configure the data sensor network of the store users based on the behavior collection requirements, calculate the behavior coverage coefficient and user interference coefficient of the data sensor network, optimize the data sensor network based on the behavior coverage coefficient, user interference coefficient and preset non-sensing data collection equipment, and obtain a target data sensor network;

[0218] Using the target data sensor network, collecting store behavior data of the offline stores, wherein the store behavior data includes video behavior data and audio behavior data;

[0219] Establishing a video motion capture function of the video behavior data, and based on the video behavior data, using the video motion capture function to capture the video behavior features of the store user, wherein the video behavior features include user posture, user running trajectory, and user emotion;

[0220] Based on the audio behavior data, the audio behavior features of the store users are extracted, wherein the audio behavior features include user audio content, user audio intonation, and user audio intensity; based on the historical store behavior data, a GRU timing network model of the store users is constructed, and a timing attention mechanism of the store users is defined; based on the video behavior features and the audio behavior features, the target user behavior of the store users is analyzed using the GRU timing network model and the timing attention mechanism.

[0221] Specifically, the specific implementation method of the processor for the above instructions can refer to the description of the relevant steps in the corresponding embodiment of the accompanying drawings, which will not be repeated here.

[0222] Furthermore, if the module / unit integrated in the electronic device is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. The computer-readable storage medium can be volatile or non-volatile. For example, the computer-readable medium can include: any entity or system that can carry the computer program code, a recording medium, a USB flash drive, a mobile hard disk, a magnetic disk, an optical disk, a computer memory, and a read-only memory (ROM).

[0223] The present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor of an electronic device, the computer program can implement:

[0224] Acquire store scene characteristics and historical store behavior data of offline stores, analyze store elements of the offline stores based on the store scene characteristics, analyze the behavior relationship between the store elements and the corresponding store users of the offline stores based on the historical store behavior data, and determine the potential user behavior of the store users based on the behavior relationship;

[0225] Analyze the behavior collection requirements of the potential user behaviors, configure the data sensor network of the store users based on the behavior collection requirements, calculate the behavior coverage coefficient and user interference coefficient of the data sensor network, optimize the data sensor network based on the behavior coverage coefficient, user interference coefficient and preset non-sensing data collection equipment, and obtain a target data sensor network;

[0226] Using the target data sensor network, collecting store behavior data of the offline stores, wherein the store behavior data includes video behavior data and audio behavior data;

[0227] Establishing a video motion capture function of the video behavior data, and based on the video behavior data, using the video motion capture function to capture the video behavior features of the store user, wherein the video behavior features include user posture, user running trajectory, and user emotion;

[0228] Based on the audio behavior data, the audio behavior features of the store users are extracted, wherein the audio behavior features include user audio content, user audio intonation, and user audio intensity; based on the historical store behavior data, a GRU timing network model of the store users is constructed, and a timing attention mechanism of the store users is defined; based on the video behavior features and the audio behavior features, the target user behavior of the store users is analyzed using the GRU timing network model and the timing attention mechanism.

[0229] In the several embodiments provided by the present invention, it should be understood that the disclosed devices, systems and methods can be implemented in other ways. For example, the system embodiments described above are only schematic, for example, the division of the modules is only a logical function division, and there may be other division methods in actual implementation.

[0230] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0231] In addition, each functional module in each embodiment of the present invention may be integrated into one processing unit, each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of hardware plus software functional modules.

[0232] It is obvious to those skilled in the art that the present invention is not limited to the details of the above exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0233] Therefore, no matter from which point of view, the embodiments should be regarded as illustrative and non-restrictive, and the scope of the present invention is limited by the appended claims rather than the above description, so it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims are included in the present invention. Any attached figure mark in the claims should not be regarded as limiting the claims involved.

[0234] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Artificial Intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0235] In addition, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or systems stated in a system claim can also be implemented by one unit or system through software or hardware. The words first, second, etc. are used to indicate names, and do not indicate any particular order.

Claims

1. A method for analyzing user behavior of offline stores based on a time series network, characterized in that: The method comprises: Acquire store scene characteristics and historical store behavior data of offline stores, analyze store elements of the offline stores based on the store scene characteristics, analyze the behavior relationship between the store elements and the corresponding store users of the offline stores based on the historical store behavior data, and determine the potential user behavior of the store users based on the behavior relationship; Analyze the behavior collection requirements of the potential user behavior, configure the data sensor network of the store users based on the behavior collection requirements, calculate the behavior coverage coefficient and user interference coefficient of the data sensor network, optimize the data sensor network based on the behavior coverage coefficient, user interference coefficient and preset non-sensing data collection equipment, and obtain the target data sensor network, wherein, configuring the data sensor network of the store users based on the behavior collection requirements includes: marking the user interaction points of the store users based on the behavior collection requirements, determining the behavior data collection equipment of the store users according to the behavior collection requirements and the user interaction points, and establishing the behavior data collection equipment Prepare a device layout at the user interaction point, identify the device network port of the behavior data collection device, build the behavior network structure of the store user based on the device network port and the device layout, and configure the data sensor network of the store user through the behavior network structure, wherein the calculation of the behavior coverage coefficient and the user interference coefficient of the data sensor network includes: identifying the device parameters of the data sensor network corresponding to the user interaction point, analyzing the device coverage area of ​​the user interaction point based on the device parameters, defining the coverage weight of the user interaction point, and calculating the behavior coverage coefficient of the data sensor network according to the coverage weight and the device coverage area using the following formula: in, represents the behavioral coverage coefficient of the data sensor network, represents the number of user interaction points corresponding to the data sensor network, Represents the data sensor network The coverage weight of user interaction points, Represents the data sensor network The device coverage area of ​​each user interaction point, Represents the data sensor network The method comprises the following steps: defining a reference area of ​​a user interaction point, defining a user interference factor of the data sensor network, determining a quantized value of the interference factor of the user interference factor, and calculating a user interference coefficient of the data sensor network based on the quantized value of the interference factor of the user interference factor, wherein the calculation of the user interference coefficient of the data sensor network based on the quantized value of the interference factor of the user interference factor includes: defining an interference factor weight of the user interference factor, analyzing an interaction point interference weight of a user interaction point corresponding to the user interference factor, and calculating the user interference coefficient of the data sensor network based on the interference factor weight, the quantized value of the interference factor, and the interaction point interference weight using the following formula: in, User interference coefficient of data sensor network, represents the number of user interaction points corresponding to the data sensor network, Represents the data sensor network The interaction point interference weight of each user interaction point, Represents the data sensor network User interaction point The interference factor weight of each user’s interference factor, Represents the data sensor network User interaction point The interference factor quantization value of the user interference factor, Represents the data sensor network The number of user interference factors per user interaction point; Using the target data sensor network, collecting store behavior data of the offline stores, wherein the store behavior data includes video behavior data and audio behavior data; Establishing a video motion capture function of the video behavior data, and based on the video behavior data, using the video motion capture function to capture the video behavior features of the store user, wherein the video behavior features include user posture, user running trajectory, and user emotion; Based on the audio behavior data, the audio behavior features of the store users are extracted, wherein the audio behavior features include user audio content, user audio intonation, and user audio intensity; based on the historical store behavior data, a GRU timing network model of the store users is constructed, and a timing attention mechanism of the store users is defined; based on the video behavior features and the audio behavior features, the target user behavior of the store users is analyzed using the GRU timing network model and the timing attention mechanism.

2. The method for analyzing user behavior of offline stores based on a time series network as claimed in claim 1, characterized in that: The analyzing, based on the historical store behavior data, the behavior relationship between the store elements and the store users corresponding to the offline stores includes: Obtaining element-related data of the store element; Time-synchronize the element-related data and the historical store behavior data to obtain a synchronized data set; Determining a user behavior type of the synchronized data set; Based on the user behavior type and the synchronized data set, construct a behavior interaction graph of the store elements; According to the behavior interaction graph, the behavior relationship between the store elements and the corresponding store users of the offline stores is analyzed.

3. The method for analyzing user behavior of offline stores based on a time series network as claimed in claim 2, characterized in that: The method of capturing the video behavior characteristics of the store user based on the video behavior data using the video motion capture function includes: Extracting the subject of the video behavior data; Constructing a three-dimensional heat map of the subject; Based on the three-dimensional heat map, the key point coordinates of the store user's body are identified using the key point recognition sub-function corresponding to the video motion capture function, wherein the key point recognition sub-function: in, Represents key points of the human body The coordinates of the key points of the human body, Indicates search The largest value, Represents the horizontal coordinate value of the three-dimensional heat map. Represents the vertical coordinate value of the three-dimensional heat map. Represents the index of the corresponding row of the three-dimensional heat map, Indicates the index of the corresponding column of the three-dimensional heat map, Represents key points of the human body , Indicates the size of the domain, Represents a three-dimensional heat map; Marking the key points of the store user's body according to the coordinates of the key points of the body; Analyzing the user posture of the store user based on the human body key points; Marking the key point coordinates of the continuous frames of the key points of the human body through the video behavior data; Connecting the key point coordinates of the consecutive frames to obtain the user running trajectory of the store user; Extracting a face image from the video behavior data; By using the facial image, analyzing the user emotion of the store user using the emotion classification function corresponding to the video motion capture function; Based on the user posture, the user running trajectory and the user emotion, the video behavior characteristics of the store user are determined.

4. The method for analyzing user behavior of offline stores based on a time series network as claimed in claim 3, characterized in that: The analyzing the user emotions of the store user by using the video motion capture function and the emotion classification function corresponding to the face image includes: De-noising the face image to obtain a de-noised face image; Correcting the denoised face image to obtain a corrected face image; Analyzing the feature vector of the corrected face image; Defining a user emotion type of the store user; Based on the feature vector and the user emotion type, the emotion classification function is used to analyze the user emotion probability of the store user: in, Represented in the feature vector User emotion type The probability of user emotions, Represents the feature vector , Indicates the user's emotion type , represents the emotion classification function, represents the weight matrix, represents the bias term, represents the transpose of the weight matrix; The user emotion of the store user is determined according to the user emotion probability.

5. The method for analyzing user behavior of offline stores based on a time series network as claimed in claim 4, characterized in that: The extracting the audio behavior features of the store user based on the audio behavior data includes: Remove background noise from the audio behavior data to obtain a de-noised audio signal; Converting the de-noised audio signal into a signal text; Extracting keywords from the signal text, and analyzing user audio content of the store user through the keywords; Segmenting the denoised audio signal into individual phonemes; analyzing frequency changes of the single phoneme; Analyzing the user audio intonation of the store user according to the frequency change; Calculating the audio energy of the denoised audio signal; Based on the audio energy, determining a user audio intensity of the store user; The audio behavior characteristics of the store user are determined based on the user audio content, the user audio tone, and the user audio intensity.

6. The method for analyzing user behavior of offline stores based on a time series network as claimed in claim 5, characterized in that: The step of constructing a GRU temporal network model of the store users based on the historical store behavior data includes: Dividing the historical store behavior data into a training set and a validation set; Define the GRU temporal network model architecture of the store user, wherein the GRU temporal network model architecture includes an input layer, a GRU layer, an output layer, and an activation function; Using the training set, training the GRU temporal network model architecture to obtain a trained GRU temporal network model; Using the validation set, the performance of the trained GRU temporal network model is verified to obtain the model performance; When the model performance meets the preset model performance threshold, the trained GRU timing network model is used as the GRU timing network model of the store user.

7. A user behavior analysis system for offline stores based on a time series network, characterized in that: The system is used to execute the method for analyzing user behavior of offline stores based on a time series network as claimed in any one of claims 1 to 6, and the system comprises: A potential user behavior analysis module is used to obtain store scene characteristics and historical store behavior data of offline stores, analyze store elements of the offline stores based on the store scene characteristics, analyze the behavior relationship between the store elements and the corresponding store users of the offline stores based on the historical store behavior data, and determine the potential user behavior of the store users based on the behavior relationship; A data sensor network construction module is used to analyze the behavior collection requirements of the potential user behaviors, configure the data sensor network of the store users based on the behavior collection requirements, calculate the behavior coverage coefficient and user interference coefficient of the data sensor network, optimize the data sensor network based on the behavior coverage coefficient, user interference coefficient and preset non-sensing data collection equipment, and obtain a target data sensor network; A behavior data collection module, used to collect store behavior data of the offline stores using the target data sensor network, wherein the store behavior data includes video behavior data and audio behavior data; A video behavior feature analysis module, used to establish a video motion capture function of the video behavior data, and based on the video behavior data, use the video motion capture function to capture the video behavior features of the store user, wherein the video behavior features include user posture, user running trajectory, and user emotion; A user behavior analysis module is used to extract the audio behavior features of the store users based on the audio behavior data, wherein the audio behavior features include user audio content, user audio intonation and user audio intensity; based on the historical store behavior data, a GRU timing network model of the store users is constructed, and a timing attention mechanism of the store users is defined; based on the video behavior features and the audio behavior features, the target user behavior of the store users is analyzed using the GRU timing network model and the timing attention mechanism.

Images