An information pushing method and system based on big data matching and GPS positioning

Abstract

The application provides a kind of information push method and system based on big data matching and GPS positioning, wherein, the application is processed by receiving position data, position data is processed using Kalman filtering algorithm, combined with time stamp and velocity vector information, trajectory prediction algorithm is applied to predict movement trend, and potential interest area list is generated;Using spatial clustering analysis algorithm, the historical position data of the user is analyzed, the behavior pattern and the preferred place type of the user are identified, and the user interest model is constructed according to it;The dynamic adaptive geofencing system is configured by using the user interest model;Weighted K nearest neighbor algorithm is used to sort service providing points, and multi-modal fusion algorithm is used to combine geographic location information in service recommendation list with other sensor data to generate customized information package;The technical scheme provided by the application improves the accuracy of position information, enhances the accuracy and timeliness of prediction, and enhances personalized service recommendation.

Description

TECHNICAL FIELD

[0001] The embodiment of the application relates to the GPS technical field, and particularly relates to an information pushing method based on big data matching and GPS positioning. BACKGROUND

[0002] With the popularity of mobile Internet and smart devices, users' demand for personalized information services is increasing. In modern urban life, people rely on mobile devices such as smartphones to obtain real-time traffic information, surrounding service recommendations, and personalized advertisement pushing, etc.

[0003] Although the existing information pushing methods have achieved location-based service recommendations to some extent, they still have some significant deficiencies. First, lack of precision, most existing solutions only rely on simple geographic location information for service recommendation, failing to fully utilize users' behavior patterns and preferred location types, resulting in inaccurate recommendation results that cannot fully reflect users' actual needs. Second, insufficient responsiveness, traditional geofencing technology usually adopts fixed boundary settings that cannot be flexibly adjusted according to users' real-time location and behavior patterns, which may miss important service providers or incorrectly include irrelevant areas when users move quickly or are in complex environments, reducing the effectiveness and relevance of service recommendations. Third, limited data processing capacity, existing systems often struggle to efficiently process high-frequency GPS coordinate updates, especially when facing simultaneous requests from a large number of users, which may lead to delays or data loss. Fourth, insufficient multi-source data fusion, existing solutions rarely effectively integrate geographic location information with other sensor data (such as images, sounds, temperature, humidity, etc.), limiting the richness and accuracy of recommended content. Finally, low degree of personalization, due to the failure to deeply mine users' long-term behavior patterns and immediate needs, existing systems' recommendation results often lack personalization, resulting in poor user experience. SUMMARY

[0004] The embodiment of the application provides an information pushing method based on big data matching and GPS positioning, which solves the problems of lack of precision, insufficient responsiveness, limited data processing capacity, insufficient multi-source data fusion, and low degree of personalization in the prior art.

[0005] In a first aspect, the embodiment of the application provides an information pushing method based on big data matching and GPS positioning, comprising:

[0006] receiving real-time data streams from different sources, the real-time data streams containing structured data and unstructured data;

[0007] Real-time receiving location data from high-frequency GPS coordinate updates in a user mobile device, and processing the location data using a Kalman filtering algorithm to obtain real-time location information;

[0008] According to the real-time location information, combining timestamp and speed vector information, applying a trajectory prediction algorithm to predict the user's moving trend, and generating a potential interest area list;

[0009] Using a spatial clustering analysis algorithm to analyze the potential interest area list and the user's historical location data, identifying the user's behavior pattern and preferred place type, and constructing a user interest model according to the user's behavior pattern and preferred place type;

[0010] Using the user interest model to configure a dynamic adaptive geofencing system, which is used to automatically adjust the fence boundary according to the user's real-time location, behavior pattern and surrounding environment changes, to generate a service provider point set;

[0011] Using a weighted K-nearest neighbor algorithm to sort multiple service provider points in the service provider point set to obtain a service recommendation list, and using a multi-modal fusion algorithm to combine geographic location information in the service recommendation list with other sensor data to generate a customized information package.

[0012] Optionally, using a spatial clustering analysis algorithm to analyze the potential interest area list and the user's historical location data, identifying the user's behavior pattern and preferred place type, and constructing a user interest model according to the user's behavior pattern and preferred place type, including:

[0013] Using a spatiotemporal consistency checking algorithm to denoise the user's historical location data, removing noise points due to GPS signal drift, and filling missing values through a time series filling algorithm, and unifying timestamp format to obtain complete and consistent user location records;

[0014] Based on the user location records, combining timestamp information, extracting the user's access behavior, identifying the user's stay records at different interest points and arrival time and departure time, and generating the user's access behavior sequence;

[0015] According to the user access behavior sequence, combining the potential interest area list, using a graph matching algorithm to convert the user's access behavior sequence into a weighted directed graph, where nodes represent interest points and edge weights represent access frequency and stay time, through the weighted directed graph, establishing the association between the user access behavior sequence and the potential interest area list, and calculating the access probability matrix of each interest point;

[0016] Using the access probability matrix, a hierarchical density peak clustering algorithm is adopted, combined with the geographic fence technology, to calculate the geographic location, access frequency and dwell time, group the interest points visited by the user, automatically adjust the clustering parameters through a dynamic threshold adjustment mechanism, and obtain the clustering results reflecting the user's frequently visited places and behavior patterns;

[0017] Based on the clustering results, a hidden Markov model is applied to analyze the mobile state transition law of the user in different time periods, identify the user's behavior patterns, and perform semantic clustering on different types of interest points to identify preferred place types;

[0018] Based on the user behavior patterns and preferred place types, the geographic distance, access frequency, dwell time and interest point theme classification are comprehensively calculated, a comprehensive score is calculated for each interest point, and a multi-dimensional user interest model is constructed.

[0019] Optionally, according to the user access behavior sequence, combined with the list of potential interest areas, a graph matching algorithm is used to convert the user's access behavior sequence into a weighted directed graph, where the nodes represent interest points and the edge weights represent access frequency and dwell time. Through the weighted directed graph, the association between the user access behavior sequence and the list of potential interest areas is established, and the access probability matrix of each interest point is calculated, including:

[0020] Using a spatiotemporal consistency checking algorithm, the movement path in the user access behavior sequence is processed for time and space continuity to obtain an undirected graph representing the user's movement pattern, where the nodes represent different interest points.

[0021] According to the dwell time and access frequency of the user at each interest point, combined with the edge weights optimized through the machine learning model, the undirected graph is weighted and directed to generate a weighted directed graph reflecting the time sequence of the user's access to the interest points.

[0022] Based on the weighted directed graph, the user access behavior sequence is compared with the interest point information in the list of potential interest areas, and a context awareness mechanism is introduced to establish the association between the two, obtaining the relevance of the adjusted interest points.

[0023] For each interest point, based on the relevance of the adjusted interest points, a Bayesian network is used for probability reasoning processing to calculate the probability of the interest point being accessed, forming an access probability matrix.

[0024] Optionally, according to the dwell time and access frequency of the user at each interest point, combined with the edge weights optimized through the machine learning model, the undirected graph is weighted and directed to generate a weighted directed graph reflecting the time sequence of the user's access to the interest points, including:

[0025] The edge weight is optimized by using a machine learning model trained based on a large amount of historical access data, to generate an optimized edge weight that can dynamically adapt to the change mode of user behavior;

[0026] According to the optimized edge weight and the time stamp information of the user access interest point, the direction of the edge between each pair of interest points in the undirected graph is determined to obtain a direction definition reflecting the actual access time sequence of the user;

[0027] The undirected graph is converted by using the direction definition to generate an initial weighted directed graph;

[0028] Based on an anomaly detection algorithm, the paths in the weighted directed graph that do not conform to the user's regular behavior mode are identified and excluded to obtain an optimized weighted directed graph that can more truly represent the user's daily movement mode.

[0029] Optionally, according to the real-time location information, combined with the time stamp and speed vector information, a trajectory prediction algorithm is applied to predict the movement trend of the user to generate a potential interest area list, including:

[0030] The real-time location information processed by the Kalman filtering algorithm is combined with the time stamp and speed vector information to reconstruct the historical movement path of the user to obtain a continuous spatiotemporal trajectory. Based on the continuous spatiotemporal trajectory, a multi-step recurrent neural network is used to perform short-term and medium-long-term prediction processing on the future movement path of the user to obtain multiple predicted movement trend results;

[0031] External environment data is introduced, and a context perception model is used to adjust the movement trend result to obtain an optimized movement trend prediction;

[0032] According to the optimized movement trend prediction, the geographical position reached by the user is predicted in combination with the interest point database in the geographic information system to generate a preliminary interest area list;

[0033] A spatiotemporal density clustering algorithm is applied to optimize the preliminary interest area list based on the geographical distribution characteristics near the predicted path to obtain a clustered and optimized interest area list;

[0034] A personalized preference model is applied to filter the clustered and optimized interest area list to preferentially recommend interest points that meet the user's preferences to generate a personalized potential interest area list.

[0035] Optionally, a dynamic adaptive geographic fence system is configured by using the user interest model, and the dynamic adaptive geographic fence system is used to automatically adjust the fence boundary according to the real-time location, behavior mode and surrounding environment change of the user to generate a service providing point set, including:

[0036] The user interest model is used for comprehensive analysis and processing of real-time location information of the user, behavior patterns of the user and preferred location types, to determine the behavior state and potential destination of the current user.

[0037] According to the real-time location information and movement trend of the user, in combination with timestamp information and speed vector information, a space-time prediction algorithm is applied to estimate the possible arrival location, to obtain a set of estimated locations.

[0038] In combination with the behavior patterns of the user, preferred location types and environmental influence factors, an initial boundary of the geofence is set and processed through a machine learning optimization algorithm, to obtain an initial geofence boundary.

[0039] A dynamic threshold adjustment mechanism is introduced, and the initial geofence boundary is automatically adjusted according to real-time feedback and new location data of the user, to generate an optimized geofence boundary.

[0040] A context-aware recommendation system is applied, in combination with real-time needs of the user, historical preferences and actual availability of service providing points, to screen and sort the service providing points in the optimized geofence, to generate an optimized set of service providing points.

[0041] Optionally, a weighted K-nearest neighbor algorithm is used to sort the multiple service providing points in the set of service providing points, to obtain a service recommendation list, and a multi-modal fusion algorithm is used to combine the geographic location information in the service recommendation list with other sensor data, to generate a customized information package, including:

[0042] A weighted K-nearest neighbor algorithm is used to comprehensively evaluate and sort the multiple service providing points in the set of service providing points, based on the user interest model, real-time location information and geographic distance, access frequency, stay time and score of the service providing points, to obtain a personalized service recommendation list.

[0043] A context-aware module is introduced, in combination with external environmental data, to dynamically adjust the personalized service recommendation list, to generate an optimized service recommendation list.

[0044] Based on the optimized service recommendation list, a multi-modal fusion algorithm is applied to integrate data from different sources, to perform multi-dimensional feature description processing on each recommended service providing point, to generate a service information card.

[0045] Natural language processing technology is used to perform semantic understanding and sentiment analysis processing on the text content in the service information card, to extract key information and optimize the expression method, to obtain a content display form.

[0046] The content display form is integrated into a comprehensive information structure, and a customized information package is generated according to the behavior patterns of the user.

[0047] In a second aspect, an embodiment of the present application provides an information pushing system based on big data matching and GPS positioning, comprising:

[0048] a prediction module configured to predict a moving trend of a user according to real-time position information, in combination with time stamp and speed vector information, by applying a trajectory prediction algorithm, and generate a list of potential interest areas;

[0049] a recognition module configured to analyze the list of potential interest areas and historical position data of the user by using a spatial clustering analysis algorithm, identify a behavior pattern and a preferred place type of the user, and construct a user interest model according to the behavior pattern and the preferred place type of the user;

[0050] a configuration module configured to configure a dynamic adaptive geofencing system by using the user interest model, the dynamic adaptive geofencing system being configured to automatically adjust a fence boundary according to a real-time position, a behavior pattern and a surrounding environment change of the user, so as to generate a set of service providing points;

[0051] a generation module configured to sort a plurality of service providing points in the set of service providing points by using a weighted K-Nearest Neighbor algorithm, obtain a service recommendation list, and combine geographic position information in the service recommendation list with other sensor data by using a multi-modal fusion algorithm, so as to generate a customized information package.

[0052] In a third aspect, an embodiment of the present application provides a computing device, comprising a processor and a memory, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute the information pushing method based on big data matching and GPS positioning according to any one of the first aspect.

[0053] In a fourth aspect, an embodiment of the present application provides a computer storage medium, which stores computer program instructions, and the computer program instructions are executed by a processor to implement the information pushing method based on big data matching and GPS positioning according to any one of the first aspect.

[0054] In the embodiment of the application, real-time position data from high-frequency GPS coordinate updates in a user's mobile device is received and processed using a Kalman filter algorithm to obtain real-time position information; according to the real-time position information, combined with timestamp and speed vector information, a trajectory prediction algorithm is applied to predict the user's movement trend to generate a potential interest area list; spatial clustering analysis algorithm is used to analyze the potential interest area list and the user's historical position data to identify the user's behavior pattern and preferred place type, and a user interest model is constructed according to the user's behavior pattern and preferred place type; the dynamic adaptive geofencing system is configured using the user interest model, which is used to automatically adjust the fence boundary according to the user's real-time position, behavior pattern and surrounding environment changes to generate a service provider point set; a weighted K-nearest neighbor algorithm is used to sort multiple service provider points in the service provider point set to obtain a service recommendation list, and a multi-modal fusion algorithm is used to combine the geographic location information in the service recommendation list with other sensor data to generate a customized information package. The technical solution provided by the application improves the accuracy of position information, enhances the accuracy and timeliness of prediction, enhances personalized service recommendation, and can dynamically respond to environmental changes, and multi-source data fusion is used to improve the recommendation quality;

[0055] Further, a space-time consistency checking algorithm is used to denoise the user's historical position data, remove noise points due to GPS signal drift, and fill in missing values through a time series filling algorithm, unify the timestamp format, and obtain complete and consistent user position records. This process ensures the quality and completeness of the data used for analysis, laying a reliable foundation for subsequent analysis;

[0056] Based on the user's location records, combined with timestamp information, the user's access behavior is extracted, the user's stay records at different interest points and arrival time and departure time are identified, and the user's access behavior sequence is generated. This helps to better understand the user's daily activity rules, so as to better meet their individual needs;

[0057] According to the user's access behavior sequence, combined with the potential interest area list, a graph matching algorithm is used to convert the user's access behavior sequence into a weighted directed graph, where the nodes represent interest points and the edge weights represent access frequency and stay time. Through this conversion, the association between user behavior patterns and interest areas can be intuitively displayed, providing strong support for subsequent analysis;

[0058] Through the weighted directed graph, the association between the user's access behavior sequence and the potential interest area list is established, and the access probability matrix of each interest point is calculated. This matrix can quantify the user's access probability to each interest point, helping the system to more accurately predict the user's future behavior and optimize service recommendation;

[0059] These and other aspects of the present application will become more apparent from the following description. BRIEF DESCRIPTION OF DRAWINGS

[0060] In order to clearly illustrate the technical solutions of the embodiments of the present application or the prior art, the following will briefly introduce the drawings required by the embodiments or the prior art description. Obviously, the drawings described below are only some embodiments of the present application, and for those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative labor.

[0061] Figure 1 A flow chart of an information pushing method based on big data matching and GPS positioning provided by an embodiment of the present application;

[0062] Figure 2 A structural schematic diagram of an information pushing system based on big data matching and GPS positioning provided by an embodiment of the present application;

[0063] Figure 3 A structural schematic diagram of a computing device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0064] In order to make the technical personnel in the art better understand the present application scheme, the following will combine the drawings in the embodiments of the present application, and clearly and completely describe the technical solutions in the embodiments of the present application.

[0065] In some of the descriptions in the specification and claims of the present application and the above-mentioned drawings, a plurality of operations appearing in a specific order are included, but it should be clearly understood that these operations can be executed or in parallel without the order appearing in this text. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes can include more or less operations, and these operations can be executed in sequence or in parallel. It should be noted that the "first", "second", etc. in this text are used to distinguish different messages, devices, modules, etc., and do not represent the order of sequence, nor do "first" and "second" represent different types.

[0066] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0067] Figure 1A flow chart of a big data processing method for implementing hybrid data analysis is provided for an embodiment of the present application, as shown in Figure 1 The method comprises the following steps:

[0068] Step 101: Real-time receiving of position data from high-frequency GPS coordinate updates in a user mobile device, and processing of the position data by using a Kalman filtering algorithm to obtain real-time position information;

[0069] In this step, high-frequency GPS coordinate updates refer to the acquisition of user geographic position information by the mobile device at a high frequency (such as once per second or several seconds);

[0070] The Kalman filtering algorithm refers to a recursive optimal estimation algorithm, which is often used to estimate the state of a process from a series of incomplete and noisy measurements;

[0071] Real-time receiving of position data from high-frequency GPS coordinate updates in a user mobile device, and processing of the position data by using a Kalman filtering algorithm to reduce errors caused by factors such as signal interference and multipath effect, to obtain more accurate and stable real-time position information.

[0072] Step 102: According to the real-time position information, combining timestamp and speed vector information, applying a trajectory prediction algorithm to predict the moving trend of the user, and generating a list of potential interest areas;

[0073] In this step, the timestamp refers to data recording the exact time of an event; the speed vector information refers to information including the speed and direction of user movement; the trajectory prediction algorithm refers to an algorithm for predicting the possible path in the future by analyzing historical trajectory data;

[0074] According to the real-time position information obtained in step 101, combining timestamp and speed vector information, and applying a trajectory prediction algorithm, the algorithm predicts the moving trend of the user, identifies the places that the user is likely to visit in the future, and generates a list of potential interest areas, which helps to prepare relevant service recommendations in advance.

[0075] Step 103: Using a spatial clustering analysis algorithm to analyze the list of potential interest areas and the historical position data of the user, identifying the behavior patterns and preferred place types of the user, and constructing a user interest model according to the behavior patterns and preferred place types of the user;

[0076] In this step, the spatial clustering analysis algorithm refers to an algorithm for grouping points in geographic space according to certain similarity criteria; the behavior pattern refers to the behavior characteristics of the user at different times and places, such as frequently visited places, activity times, etc.; the preferred place type refers to the preferred place category of the user, such as restaurants, parks, shopping malls, etc.;

[0077] The spatial clustering analysis algorithm is used to analyze the list of potential interest areas and the historical location data of the user, identify the behavior patterns and preferred location types of the user, and construct a user interest model based on the information, which can reflect the daily activity rules and personal preferences of the user and provide a basis for personalized service recommendation.

[0078] Step 104: configuring a dynamic adaptive geofencing system using the user interest model, the dynamic adaptive geofencing system being used to automatically adjust the fence boundary according to the real-time location, behavior pattern and surrounding environment change of the user to generate a service providing point set;

[0079] In this step, the dynamic adaptive geofencing system refers to a virtual fence system that can automatically adjust the boundary according to the real-time location, behavior pattern and surrounding environment change of the user; the service providing point set refers to a group of locations or facilities that can provide services for the user, such as shops, restaurants, scenic spots, etc.

[0080] The dynamic adaptive geofencing system is configured using the user interest model, which automatically adjusts the fence boundary according to the real-time location, behavior pattern and surrounding environment change of the user, ensures that the service providing points within the fence are highly relevant to the current needs of the user, and generates a service providing point set that is the basis for subsequent service recommendation, ensuring the timeliness and accuracy of the recommendation results.

[0081] Step 105: using a weighted K-nearest neighbor algorithm to sort multiple service providing points in the service providing point set to obtain a service recommendation list, and using a multi-modal fusion algorithm to combine the geographic location information in the service recommendation list with other sensor data to generate a customized information package;

[0082] In this step, the dynamic adaptive geofencing system refers to a virtual fence system that can automatically adjust the boundary according to the real-time location, behavior pattern and surrounding environment change of the user; the service providing point set refers to a group of locations or facilities that can provide services for the user, such as shops, restaurants, scenic spots, etc.

[0083] The dynamic adaptive geofencing system is configured using the user interest model, which automatically adjusts the fence boundary according to the real-time location, behavior pattern and surrounding environment change of the user, ensures that the service providing points within the fence are highly relevant to the current needs of the user, and generates a service providing point set that is the basis for subsequent service recommendation, ensuring the timeliness and accuracy of the recommendation results.

[0084] Assuming that a tourist, Xiao Li, is visiting a famous scenic spot, he opens the smart travel APP on his phone, hoping to obtain information about nearby restaurants, shops, and other services. The system receives real-time location data from Xiao Li's mobile device, which updates high-frequency GPS coordinates, and processes these data using Kalman filtering algorithms to remove noise interference and obtain accurate real-time location information (such as longitude and latitude). Next, the system uses trajectory prediction algorithms to predict Xiao Li's movement trends based on real-time location information, time stamps, and speed vector information, generating a list of potential interest areas, such as nearby major scenic spots, restaurants, shopping malls, etc. Then, using spatial clustering analysis algorithms, the system analyzes the list of potential interest areas and Xiao Li's historical location data to identify his behavior patterns (such as a preference for visiting historical and cultural scenic spots and a preference for afternoon tea period cafes) and preferred location types (such as museums, art galleries, and specialty restaurants), and builds a user interest model reflecting Xiao Li's interests and preferences.

[0085] Further, the system configures a dynamic adaptive geofencing system using the user interest model, automatically adjusts the fence boundary based on Xiao Li's real-time location, behavior patterns, and changes in the surrounding environment (such as traffic conditions and weather conditions), ensuring that the recommended service points not only meet Xiao Li's interests but also take into account the current actual situation. For example, when detecting that there is a large-scale activity nearby causing traffic congestion, the system will adjust the fence range accordingly, recommend alternative routes or more convenient service points, and then use a weighted K-nearest neighbor algorithm to sort multiple service providers in the service provider set, obtaining a service recommendation list such as nearby popular restaurants and exhibition activities. Then, using a multi-modal fusion algorithm, the system combines the geographic location information in the service recommendation list with other sensor data (such as weather forecasts and traffic flow) to generate a customized information package. The customized information package not only contains accurate service recommendations but also integrates other types of rich information such as pictures and audio introductions, enabling Xiao Li to fully understand the specific circumstances of each recommended point.

[0086] Finally, the system pushes the customized information package to Xiao Li. When Xiao Li approaches a historical museum, the system pushes detailed information about the museum and information about a special exhibition to be held. When Xiao Li's lunch time is approaching, the system recommends several local restaurants with high ratings and attaches menus and user reviews. In addition, considering that it is raining that day, the system also thoughtfully recommends some indoor activity venues. All this information is intuitively displayed to Xiao Li through the APP, helping him better plan his itinerary, significantly improving his travel experience, making his journey smoother and more enjoyable, while also increasing the exposure of service providers and user satisfaction, achieving a win-win situation. Through this complete implementation process, it is shown how the method of claim 1 can effectively improve the accuracy, relevance, and user experience of information pushing in actual applications, meeting the diverse needs of users.

[0087] Based on this, the application provides a specific embodiment, wherein the step 203 is to convert the user's access behavior sequence into a weighted directed graph by using a graph matching algorithm according to the user's access behavior sequence and in combination with the list of potential interest areas, wherein a node represents a point of interest, and an edge weight represents an access frequency and a stay time, to establish the relevance between the user's access behavior sequence and the list of potential interest areas through the weighted directed graph, and to calculate an access probability matrix of each point of interest, specifically including the following steps:

[0088] Step 210: performing denoising processing on the user's historical location data by using a spatiotemporal consistency checking algorithm, removing noise points caused by GPS signal drift, filling in missing values by using a time series filling algorithm, and unifying a timestamp format to obtain complete and consistent user location records.

[0089] In this step, the spatiotemporal consistency checking algorithm refers to an algorithm for detecting and correcting inconsistencies in time series data to ensure the consistency of time and space data; the GPS signal drift refers to the GPS positioning error caused by satellite signal interference or other factors, which is manifested as random fluctuations in location data; the time series filling algorithm refers to an algorithm for filling in missing values in time series data to ensure data continuity and integrity; and the timestamp format unification refers to converting timestamps from different sources into a unified standard format to facilitate subsequent analysis.

[0090] The spatiotemporal consistency checking algorithm is used to perform denoising processing on the user's historical location data, remove noise points caused by GPS signal drift, fill in missing values in the historical location data by using a time series filling algorithm, and unify all timestamp formats to ensure data completeness and consistency, so as to finally obtain a cleaned and standardized user location record, providing a reliable data basis for subsequent steps.

[0091] Step 211: based on the user location record, in combination with timestamp information, extracting the user's access behavior, identifying the user's stay records at different points of interest and the arrival time and departure time, and generating the user's access behavior sequence.

[0092] In this step, the user location record refers to the user's moving track data after denoising and filling; the access behavior refers to the user's stay behavior at a specific place, including the arrival time and departure time; and the point of interest (POI) refers to a place with specific significance or attraction, such as a store, a restaurant, etc.

[0093] Based on the user location record, in combination with timestamp information, the user's stay records at different points of interest and the specific arrival time and departure time are identified, and the user's access behavior sequence is generated, which describes the user's activity track in detail within a period of time and provides an important basis for subsequent analysis.

[0094] Step 212: According to the user access behavior sequence, combined with the list of potential interest areas, the user's access behavior sequence is converted into a weighted directed graph using a graph matching algorithm, where the nodes represent interest points and the edge weights represent access frequency and dwell time. Through the weighted directed graph, the association between the user access behavior sequence and the list of potential interest areas is established, and the access probability matrix of each interest point is calculated.

[0095] In this step, the graph matching algorithm refers to an algorithm for comparing the similarity of two graph structures, which is used to convert the access behavior sequence into a graph representation. The weighted directed graph refers to a graph representation method in which the nodes represent interest points and the edge weights represent access frequency and dwell time. The access probability matrix refers to a matrix that describes the probability distribution from one interest point to another.

[0096] According to the user access behavior sequence, combined with the list of potential interest areas, the user's access behavior sequence is converted into a weighted directed graph using a graph matching algorithm, where the nodes represent interest points and the edge weights represent user access frequency and dwell time. Through the weighted directed graph, the association between the user access behavior sequence and the list of potential interest areas is established, and the access probability matrix of each interest point is calculated, providing a basis for subsequent clustering analysis.

[0097] Step 213: Using the access probability matrix, a hierarchical density peak clustering algorithm is used in combination with geofencing technology to calculate geographic location, access frequency, and dwell time to group interest points visited by the user. A dynamic threshold adjustment mechanism automatically adjusts clustering parameters to obtain clustering results that reflect the user's frequently visited locations and behavior patterns.

[0098] In this step, the hierarchical density peak clustering algorithm refers to a density-based clustering algorithm that can automatically discover natural groupings in data. Geofencing technology refers to defining a virtual boundary to monitor the user's entry or exit from a specific area. The dynamic threshold adjustment mechanism refers to automatically adjusting algorithm parameters based on actual conditions to ensure the accuracy and adaptability of the clustering results.

[0099] Using the access probability matrix, a hierarchical density peak clustering algorithm is used in combination with geofencing technology to calculate geographic location, access frequency, and dwell time to group interest points visited by the user. A dynamic threshold adjustment mechanism automatically adjusts clustering parameters to ensure that the clustering results can reflect the user's frequently visited locations and behavior patterns, providing support for personalized service recommendations.

[0100] Step 214: Based on the clustering results, an HMM model is applied to analyze the user's movement state transition rules in different time periods to identify the user's behavior patterns, and a semantic clustering of different types of interest points is performed to identify preferred location types.

[0101] In this step, the hidden Markov model (HMM) refers to a statistical model used to represent a system composed of a series of possible invisible states, and there is a certain transition probability between the states; semantic clustering refers to classifying interest points with similar semantic features into a class, such as "dining", "shopping", etc.

[0102] Based on the clustering results, the hidden Markov model is applied to analyze the movement state transition law of the user in different time periods, to identify the behavior mode of the user, and at the same time, to perform semantic clustering on different types of interest points, to identify the preferred place type, to further refine the user interest model, and to improve the accuracy of personalized recommendation.

[0103] Step 215: Based on the user behavior mode and the preferred place type, the geographic distance, the access frequency, the stay time and the interest point theme classification are comprehensively calculated, a comprehensive score is calculated for each interest point, and a multi-dimensional user interest model is constructed;

[0104] In this step, the comprehensive score refers to a score calculated according to multiple dimensions (such as geographic distance, access frequency, stay time, and interest point theme classification), which is used to evaluate the interest degree of the user to a certain interest point; the multi-dimensional user interest model refers to a model constructed by comprehensively considering multiple factors, which can comprehensively reflect the user's interest and preference;

[0105] Based on the user behavior mode and the preferred place type, the geographic distance, the access frequency, the stay time and the interest point theme classification are comprehensively calculated, a comprehensive score is calculated for each interest point, and a multi-dimensional user interest model is constructed, which not only reflects the geographic location preference of the user, but also considers the access frequency, the stay time and the theme classification of the interest point, to provide highly personalized and accurate service recommendation for the user.

[0106] Based on this, the present application provides a specific embodiment, and the step 212 comprises the following steps:

[0107] Step 301: The time and space coherence checking algorithm is used to perform time and space coherence processing on the movement path in the user access behavior sequence, to obtain an undirected graph representing the movement mode of the user, wherein the nodes represent different interest points;

[0108] In this step, the spatiotemporal consistency checking algorithm refers to an algorithm for ensuring the consistency of time and space data, which can detect and correct discontinuous data points; the user access behavior sequence refers to time series data recording the user's stay time, arrival time and departure time at different interest points; the undirected graph refers to a graphical representation method in which nodes represent interest points and edges have no direction, representing the relationship between two interest points;

[0109] The spatiotemporal consistency checking algorithm is used to process the movement path in the user access behavior sequence for time and space continuity, ensuring that the time and space relationship between each location point is reasonable and consistent. Through this processing, an undirected graph representing the user's movement pattern is generated, in which the nodes represent different interest points. This graph reflects the user's movement path between different interest points, providing a structured basis for subsequent analysis.

[0110] Step 302: According to the user's stay time and access frequency at each interest point, combined with the edge weight optimized by the machine learning model, the undirected graph is weighted and directed processed to generate a weighted directed graph reflecting the time sequence of the user's access to the interest point;

[0111] In this step, the stay time refers to the length of time the user stays at an interest point; the access frequency refers to the number of times the user visits a certain interest point; the edge weight optimized by the machine learning model refers to the weight value obtained through machine learning model training, used to reflect the strength of the association between interest points;

[0112] According to the user's stay time and access frequency at each interest point, combined with the edge weight optimized by the machine learning model, the undirected graph obtained in step 301 is weighted and directed processed. Weighted processing assigns each edge a weight value representing the importance of user access frequency and stay time; directional processing determines the direction of the edge according to the actual movement sequence of the user, finally generating a weighted directed graph reflecting the time sequence of the user's access to the interest point. This process makes the user's behavior pattern more intuitively reflected in the graph structure, facilitating subsequent analysis and application.

[0113] Step 303: Based on the weighted directed graph, compare the user access behavior sequence with the interest point information in the potential interest area list, and introduce a context-aware mechanism to establish the relevance between the two, obtaining the relevance of the adjusted interest points;

[0114] In this step, the weighted directed graph refers to a graphical representation method in which nodes represent interest points and edges have direction and weight, representing access frequency and stay time; the context-aware mechanism refers to considering factors such as current time, weather conditions, public events, etc., to adjust the relevance of interest points; relevance: an indicator measuring the likelihood or importance of a user accessing a certain interest point;

[0115] Based on the weighted directed graph generated in step 302, the user access behavior sequence is compared with the interest point information in the potential interest area list, a context awareness mechanism is introduced, factors such as current time, weather conditions and special events are considered, the relevance of the interest point is adjusted to more accurately reflect the real-time needs and preferences of the user, the relevance between the user access behavior sequence and the potential interest area list is established, and the relevance of the adjusted interest point is obtained to provide a basis for subsequent probability reasoning.

[0116] Step 304: For each interest point, based on the adjusted relevance of the interest point, a probability reasoning process is performed using a Bayesian network to calculate the probability of the interest point being visited, forming a visit probability matrix.

[0117] In this step, the Bayesian network refers to a probabilistic graphical model for representing the dependency relationship between variables and supporting probability reasoning; the visit probability matrix refers to a probability distribution matrix describing the transition from one interest point to another;

[0118] For each interest point, based on the adjusted relevance of the interest point, a probability reasoning process is performed using a Bayesian network to calculate the probability of the interest point being visited, forming a visit probability matrix. This matrix not only reflects the likelihood of the user visiting each interest point, but also captures the transition probability between interest points, providing an important reference for subsequent clustering analysis and service recommendation;

[0119] Let I = {i1, i2, …, in} be the set of interest points visited by the user, T { i n } is the set of interest points visited by the user, T ( i j ) is the user's stay time at interest point i j , F ( i j ) is the visit frequency, R ( i j ) is the adjusted relevance (considering the context awareness mechanism), C(t) is the time decay factor, E ( i j ) is the external environmental influence factor (such as weather, traffic conditions, etc.), then the visit probability matrix P can be calculated by the following complex formula:

[0120]

[0121] where I = {i1, i2, …, in} is the set of interest points visited by the user, T { i n}T ( i j ) represents the set of interest points visited by the user, extracted from the user's mobile trajectory data (such as GPS records), representing a series of locations that the user has visited; ( i j ) represents the set of interest points visited by the user, extracted from the user's mobile trajectory data (such as GPS records), representing a series of locations that the user has visited; j i ( ) represents the duration of the user's stay at interest point i j ) , usually determined by the time difference between the user's entry and exit times at a certain interest point, extracted from the user's historical location data; j i ( ) represents the frequency of the user's visits to interest point i j ) , counted by the number of times the user visits a certain interest point in the historical behavior data; ( i j ) , constructed by combining the user's historical behavior patterns and preferred location types through spatial clustering analysis algorithms, considering the context-aware mechanism;

[0122] C(t) represents the time decay factor, indicating that the influence of historical access behavior on current prediction gradually decreases over time, where C(t) = exp(-λt), t is the time since the last visit, and λ is the decay rate parameter, reducing the influence of historical behavior on current prediction, making the prediction more close to the current behavior pattern;E ( i j ) , considering the influence of external environment (such as weather, traffic conditions, etc.) on the user's visits to interest points, where E ( i j ) = w1W ( i j ) +w2T r ( i j ) , W ( i j ) is the weather influence factor, such as sunny, rainy, etc., T r ( i j ) is the traffic condition influence factor, such as the degree of traffic congestion; w1 and w2 are the corresponding weight coefficients, considering the influence of external factors on user behavior, making the prediction result more consistent with the actual situation; α, β, γ, δ, η represent weight parameters, used to balance the importance of different factors, which can be optimized by machine learning models to ensure that the final access probability matrix accurately reflects the user's interests and behavior patterns;

[0123] The numerator and denominator parts are explained as follows:

[0124] exp(alpha*T(i j )+beta*F(i j )+gamma*R(i j )+delta*C(t)+eta*E(i j ))

[0125] The weighted comprehensive score of the interest point ij is calculated by combining each factor through an exponential function, enhancing the relative importance difference between different factors.

[0126]

[0127] The weighted comprehensive scores of all interest points are normalized to ensure that the sum of the probabilities of all interest points is 1, forming a probability distribution.

[0128] Wherein, the longer the residence time T(i j , the greater the user's interest in the interest point, so a higher weight is given when calculating the access probability; the higher the access frequency F(i j , the more stable the user's interest in the interest point, increasing the possibility of being recommended; the adjusted relevance R(i j ) indicates the relevance of the interest point adjusted by the context perception mechanism, reflecting the user's preference for the interest point and improving the accuracy of personalized recommendation; the time decay factor C(t) indicates the reduction of the influence of historical behavior on current prediction, making the recommendation result more close to the user's current behavior pattern; the external environment influence factor E(i j ) indicates the influence of external environment (such as weather, traffic conditions, etc.) on the behavior, making the recommendation result more in line with the actual situation.

[0129] Based on this, the present application provides a specific embodiment, wherein the step 302, according to the residence time and access frequency of the user at each interest point, combines the edge weight optimized by the machine learning model, and performs weighted and directional processing on the undirected graph to generate a weighted directed graph reflecting the time sequence of the user accessing the interest point, specifically including the following steps:

[0130] Step 401: using a machine learning model trained based on a large amount of historical access data to optimize the edge weight, generating an optimized edge weight that can dynamically adapt to the user's behavior change pattern;

[0131] In this step, the machine learning model refers to an algorithm that can automatically improve and adjust parameters through training data, and can make predictions or decisions on new data; the historical access data refers to information such as the time and frequency of the user's past visits to different points of interest, which is used to train the model; the edge weight optimization processing refers to adjusting the weight of the edge in the graph through the machine learning model, so that it more accurately reflects the user behavior pattern;

[0132] The machine learning model trained based on a large amount of historical access data is used to optimize the edge weight in the undirected graph. This model can dynamically adapt to the changing patterns of user behavior, ensuring that the edge weight not only reflects the user's current behavior, but also predicts future trends, ultimately generating optimized edge weights that can dynamically adapt to the changing patterns of user behavior, providing a more accurate data foundation for subsequent steps.

[0133] Step 402: According to the optimized edge weight and the timestamp information of the user's visit to the point of interest, the direction of each edge between each pair of points of interest in the undirected graph is determined, and the direction definition reflecting the actual visit time sequence of the user is obtained;

[0134] In this step, the timestamp information refers to the specific time recorded when the user visits each point of interest; the direction determination processing refers to assigning a direction to each edge in the graph, indicating the time sequence of the user moving from one point of interest to another;

[0135] According to the optimized edge weight and the timestamp information of the user's visit to the point of interest, the direction of each edge between each pair of points of interest in the undirected graph is determined, and the direction definition reflecting the actual visit time sequence of the user is obtained, making the graph structure more consistent with the user's movement pattern.

[0136] Step 403: Convert the undirected graph using the direction definition to generate an initial weighted directed graph;

[0137] In this step, the conversion processing refers to the process of converting the undirected graph into a directed graph, including assigning directions to edges and updating weights; the initial weighted directed graph refers to the graphical representation after the conversion processing, with directions and optimized weights, which preliminarily reflects the user's behavior pattern;

[0138] Using the direction definition determined in step 402, the undirected graph is converted into a weighted directed graph, where each edge has a clear direction and optimized weight, generating an initial weighted directed graph. This graph not only reflects the time sequence of the user's visit to the point of interest, but also embodies the importance of visit frequency and dwell time through edge weight, providing structured data support for further analysis.

[0139] Step 404: based on the anomaly detection algorithm, the paths in the weighted directed graph that do not conform to the user's regular behavior pattern are identified and excluded, and an optimized weighted directed graph that can more truly represent the user's daily movement pattern is obtained;

[0140] In this step, the anomaly detection algorithm refers to an algorithm that identifies data points or paths in the data set that do not conform to the regular pattern; the optimized weighted directed graph refers to a weighted directed graph that can more truly represent the user's daily movement pattern by removing abnormal paths;

[0141] Based on the anomaly detection algorithm, the paths in the weighted directed graph generated in step 403 that do not conform to the user's regular behavior pattern are identified and excluded, and the abnormal paths may include unusual movements or error data points caused by accidental factors, which will interfere with the accurate analysis of the user's behavior pattern. By removing these abnormal paths, an optimized weighted directed graph that can more truly represent the user's daily movement pattern is obtained, ensuring the reliability and accuracy of subsequent analysis and recommendation results.

[0142] Based on this, the present application provides a specific embodiment, wherein the step 102 comprises the following steps:

[0143] Step 501: using the real-time position information processed by the Kalman filtering algorithm, combining the timestamp and velocity vector information, reconstructing the user's historical movement path to obtain a continuous spatiotemporal trajectory, and based on the continuous spatiotemporal trajectory, using a multi-step recurrent neural network to perform short-term and medium-long-term prediction processing on the user's future movement path to obtain multiple predicted movement trend results;

[0144] In this step, the Kalman filtering algorithm refers to a recursive optimal estimation algorithm used to estimate the state of a process from a series of incomplete and noisy measurements; the real-time position information refers to the user's current position data obtained by GPS or other means; the timestamp and velocity vector information refer to the time of the event and the speed and direction of the user's movement; the multi-step recurrent neural network (RNN) refers to a deep learning model that can process sequence data and make predictions for multiple future time points;

[0145] The real-time position information processed by the Kalman filtering algorithm is combined with timestamp and speed vector information to reconstruct the historical moving path of the user, and a continuous space-time trajectory is obtained. Based on the continuous space-time trajectory, a multi-step recursive neural network (such as LSTM or GRU) is used to perform short-term and medium-long term prediction processing on the future moving path of the user. This process generates multiple predicted moving trend results, covering the possible moving paths of the user in different time periods, providing a rich prediction basis for subsequent steps.

[0146] Step 502: Introduce external environment data, and adjust the moving trend result by using a context perception model to obtain an optimized moving trend prediction.

[0147] In this step, external environment data refers to external factors that affect user behavior, including traffic flow, weather conditions, public events, etc. The context perception model refers to a model that considers various factors in the current situation to adjust the prediction result.

[0148] Introducing external environment data, and adjusting the moving trend result obtained in step 501 by using a context perception model, the context perception model will dynamically adjust the prediction result according to the current time, weather conditions, traffic conditions, etc. to ensure that it is more in line with the actual situation, and finally obtain an optimized moving trend prediction, improving the accuracy and practicality of the prediction result.

[0149] Step 503: Based on the optimized moving trend prediction, combine the point of interest database in the geographic information system to predict the geographic location reached by the user, and generate a preliminary interest area list.

[0150] In this step, the geographic information system (GIS) refers to a system used to capture, store, manipulate, analyze, manage, and display all types of geographic data. The point of interest database refers to an information library containing various points of interest (POI), such as restaurants, stores, scenic spots, etc.

[0151] Based on the optimized moving trend prediction, combine the point of interest database in the geographic information system to predict the geographic location that the user may reach in the future, analyze these geographic locations, and identify potential points of interest to generate a preliminary interest area list. This step lays the foundation for subsequent point of interest screening and recommendation.

[0152] Step 504: Apply a space-time density clustering algorithm based on the geographic distribution characteristics near the predicted path to optimize the preliminary interest area list and obtain a clustered and optimized interest area list.

[0153] In this step, the spatiotemporal density clustering algorithm refers to a density clustering algorithm based on time and space dimensions, such as DBSCAN or OPTICS, used to identify dense groups of data points; the geographical distribution feature refers to the spatial distribution of interest points in a certain area, such as density, distance, etc.

[0154] By applying the spatiotemporal density clustering algorithm, the preliminary interest region list is optimized based on the geographical distribution characteristics near the predicted path. Through clustering analysis, low-probability interest points are removed, and high-probability access regions are retained, thereby obtaining a clustering-optimized interest region list. This process not only reduces unnecessary interest points, but also improves the relevance and accuracy of the recommendation results.

[0155] Step 505: Apply the personalized preference model to filter the clustering-optimized interest region list, prioritize the recommendation of interest points that meet user preferences, and generate a personalized potential interest region list.

[0156] In this step, the personalized preference model refers to a model trained from user historical behavior data through machine learning algorithms, used to predict user preferences; the personalized potential interest region list refers to the final recommended interest point list generated after filtering according to user preferences.

[0157] By applying the personalized preference model to filter the clustering-optimized interest region list, the interest points that meet the user's preferences are prioritized. Based on the user's historical behavior patterns and immediate needs, this model ensures that the recommended interest points are highly personalized and more closely meet the user's actual needs, ultimately generating a personalized potential interest region list to provide accurate and personalized service recommendations for users.

[0158] Based on this, the present application provides a specific embodiment, wherein step 105 adopts a weighted K-nearest neighbor algorithm to sort multiple service provider points in the service provider point set to obtain a service recommendation list, and utilizes a multi-modal fusion algorithm to combine geographical location information in the service recommendation list with other sensor data to generate a customized information package, specifically including the following steps:

[0159] Step 701: Utilize a weighted K-nearest neighbor algorithm to comprehensively evaluate and sort multiple service provider points in the service provider point set based on a user interest model, real-time location information, and geographical distance, access frequency, dwell time, and ratings of service provider points, to obtain a personalized service recommendation list.

[0160] In this step, the weighted K-nearest neighbor algorithm (WKNN) refers to a classification or regression algorithm based on distance measurement, where the weight of each neighbor is determined by its distance; the user interest model refers to a model constructed based on the user's behavior patterns and preferred location types, used to predict the user's interests and needs; real-time location information refers to the user's geographic location data at the current time; the service provider point set refers to a set of locations or facilities that can provide services to the user;

[0161] Using the weighted K-nearest neighbor algorithm, the user interest model, real-time location information, and multiple factors such as geographic distance, access frequency, stay time, and ratings of service provider points are used to comprehensively evaluate and rank multiple service provider points in the service provider point set. Through this multi-dimensional evaluation, a personalized service recommendation list is generated, ensuring that the recommended results not only meet the user's preferences, but also take into account the actual geographic location and characteristics of the service provider points.

[0162] Step 702: Introduce the context-aware module, combine external environment data, dynamically adjust the personalized service recommendation list, and generate an optimized service recommendation list;

[0163] In this step, the context-aware module refers to a system component that can identify and respond to environmental changes, and can consider external environment data to adjust the recommended results; external environment data refers to external factors that affect user behavior, including traffic flow, weather conditions, public events, etc.

[0164] Introducing the context-aware module, combining external environment data (such as traffic conditions, weather conditions, public events, etc.), dynamically adjusting the personalized service recommendation list generated in step 701, the context-aware module will automatically adjust the content and order of the recommendation list according to the current time, location and environmental changes, ensuring that the recommended results are more in line with the user's immediate needs and changes in the surrounding environment.

[0165] Finally, an optimized service recommendation list is generated, improving the timeliness and relevance of the recommended results.

[0166] Step 703: Based on the optimized service recommendation list, apply a multi-modal fusion algorithm to integrate data from different sources and perform multi-dimensional feature description processing on each recommended service provider point to generate a service information card.

[0167] In this step, the multi-modal fusion algorithm refers to an algorithm that integrates information from different sensors or data sources to improve decision quality; multi-dimensional feature description processing refers to describing the characteristics of an object from multiple angles, such as geographic location, image, sound, temperature and humidity, etc.

[0168] Based on the optimized service recommendation list, a multi-modal fusion algorithm is applied to integrate data from different sources (such as geographic location information, image recognition results, sound analysis, temperature and humidity sensor data) to perform multi-dimensional feature description processing on each recommended service providing point. This process generates detailed service information cards, each of which contains rich information about the service providing point, enabling users to fully understand the specific circumstances of each recommended point and make more informed choices.

[0169] Step 704: Utilize natural language processing technology to perform semantic understanding and sentiment analysis processing on the text content in the service information card, extract key information and optimize expression methods, and obtain content display forms;

[0170] In this step, natural language processing technology (NLP) refers to the technology related to the interaction between computers and human language for text content understanding and generation; semantic understanding refers to the analysis of the meaning of text content and the extraction of key information; sentiment analysis refers to the evaluation of the emotional tendency expressed in the text, such as positive, negative or neutral;

[0171] Utilizing natural language processing technology to perform semantic understanding and sentiment analysis processing on the text content in the service information card, extracting key information and optimizing expression methods, makes the information display more concise and clear, easy for users to understand, and obtains optimized content display forms, not only improving user experience, but also enhancing the effective transmission of information.

[0172] Step 705: Integrate the content display forms into a comprehensive information structure and generate customized information packages based on user behavior patterns;

[0173] In this step, the comprehensive information structure organizes different types of information into a whole to ensure the coherence and integrity of the information.

[0174] Customized information package: combining various types of information into a comprehensive information service according to user needs and preferences;

[0175] Integrating the content display forms obtained in step 704 into a comprehensive information structure ensures that all recommended content is logically clear and consistent. Based on user behavior patterns and historical preferences, customized information packages are generated, which not only contain precise service recommendations but also integrate other types of rich information to provide a comprehensive and personalized service experience for users. The final customized information package aims to provide highly personalized, information-rich and easy-to-understand content, significantly improving user satisfaction and service quality.

[0176] Figure 2 A structural diagram of a big data processing system for implementing hybrid data analysis is provided for the embodiments of the present application, as shown inFigure 2 As shown in the figure, the system comprises:

[0177] The receiving module 21 is configured to receive position data from high-frequency GPS coordinate updates in a user mobile device in real time, and process the position data using a Kalman filtering algorithm to obtain real-time position information.

[0178] The prediction module 22 is configured to predict the moving trend of the user by applying a trajectory prediction algorithm according to the real-time position information, in combination with timestamp and speed vector information, to generate a list of potential interest areas.

[0179] The identification module 23 is configured to analyze the list of potential interest areas and historical position data of the user using a spatial clustering analysis algorithm, identify the behavior pattern and preferred place type of the user, and construct a user interest model according to the behavior pattern and preferred place type of the user.

[0180] The configuration module 24 is configured to configure a dynamic adaptive geofencing system using the user interest model, wherein the dynamic adaptive geofencing system is configured to automatically adjust the fence boundary according to the real-time position, behavior pattern, and surrounding environment changes of the user to generate a set of service providing points.

[0181] The generation module 25 is configured to sort multiple service providing points in the set of service providing points using a weighted K-nearest neighbor algorithm to obtain a service recommendation list, and combine the geographic position information in the service recommendation list with other sensor data using a multi-modal fusion algorithm to generate a customized information package.

[0182] Figure 2 The information push system based on big data matching and GPS positioning can perform Figure 1 The information push method based on big data matching and GPS positioning according to the embodiments described above will not be described in detail. The specific operation of each module and unit in the information push system based on big data matching and GPS positioning described above has been described in detail in the embodiments related to the method, and will not be described in detail here.

[0183] In one possible design, Figure 2 The information push system based on big data matching and GPS positioning according to the embodiments described above can be implemented as a computing device, such as Figure 3 As shown in the figure, the computing device can include a storage component 31 and a processing component 32.

[0184] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 32.

[0185] The processing component 32 is configured to receive position data from high-frequency GPS coordinate updates in a user mobile device in real time, and process the position data using a Kalman filtering algorithm to obtain real-time position information; according to the real-time position information, in combination with timestamp and speed vector information, apply a trajectory prediction algorithm to predict the moving trend of the user, and generate a list of potential interest areas; use a spatial clustering analysis algorithm to analyze the list of potential interest areas and the historical position data of the user, identify the behavior pattern and preferred place type of the user, and construct a user interest model according to the behavior pattern and preferred place type of the user; configure a dynamic adaptive geofencing system using the user interest model, the dynamic adaptive geofencing system being configured to automatically adjust the fence boundary according to the real-time position, behavior pattern and surrounding environment change of the user, to generate a set of service providing points; use a weighted K-nearest neighbor algorithm to sort a plurality of service providing points in the set of service providing points to obtain a service recommendation list, and use a multi-modal fusion algorithm to combine the geographic position information in the service recommendation list with other sensor data to generate a customized information package.

[0186] The processing component 32 can include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component can also be one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic elements for executing the above method.

[0187] The storage component 31 is configured to store various types of data to support the operation of the terminal. The storage component can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0188] Of course, the computing device can also include other components, such as an input / output interface, a display component, a communication component, etc.

[0189] The input / output interface provides an interface between the processing component and peripheral interface modules, which can be output devices, input devices, etc.

[0190] The communication component is configured to facilitate wired or wireless communication between the computing device and other devices, etc.

[0191] The computing device can be a physical device or an elastic computing host provided by a cloud computing platform, and the computing device can be a cloud server, and the processing component and the storage component can be basic server resources rented or purchased from the cloud computing platform.

[0192] The embodiment of the present application also provides a computer storage medium storing a computer program, and the computer program can realize the above-mentioned method when being executed by a computer. Figure 1 The embodiment shown in the figure provides an information pushing method based on big data matching and GPS positioning.

[0193] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-mentioned system, device and unit can refer to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0194] The device embodiments described above are only schematic, and the units described as separate components can or can not be physically separated, and the components shown as units can or can not be physical units, that is, they can be located in one place, or can be distributed on multiple network units. Part or all of the modules can be selected to achieve the purpose of the embodiment scheme according to actual needs. Those skilled in the art can understand and implement without creative labor.

[0195] Through the description of the foregoing embodiments, those skilled in the art can clearly understand that each embodiment can be realized by means of software and a necessary general hardware platform, and of course, it can also be realized by hardware. Based on such understanding, the above technical solutions can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, etc., and includes a plurality of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the method described in each embodiment or some parts of the embodiment.

[0196] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement to some technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. An information push method based on big data matching and GPS positioning, characterized in that, include: The system receives location data updated in real time from high-frequency GPS coordinates on the user's mobile device, and processes the location data using a Kalman filter algorithm to obtain real-time location information. Based on the real-time location information, combined with timestamp and velocity vector information, a trajectory prediction algorithm is applied to predict the user's movement trend and generate a list of potential interest areas. Spatial clustering analysis algorithms are used to analyze the list of potential interest regions and the user's historical location data to identify the user's behavioral patterns and preferred location types. Based on the user's behavioral patterns and preferred location types, a user interest model is constructed. The user interest model is used to configure a dynamic adaptive geofencing system, which is used to automatically adjust the fence boundary according to the user's real-time location, behavior pattern and changes in the surrounding environment to generate a set of service provision points; The weighted K-nearest neighbor algorithm is used to sort multiple service providers in the service provider set to obtain a service recommendation list. Then, a multimodal fusion algorithm is used to combine the geographical location information in the service recommendation list with other sensor data to generate a customized information package. Spatial clustering analysis algorithms are used to analyze the list of potential interest regions and the user's historical location data to identify user behavior patterns and preferred location types. Based on these patterns and preferences, a user interest model is constructed, including: The spatiotemporal consistency check algorithm is used to denoise the user's historical location data, removing noise points caused by GPS signal drift. Missing values ​​are filled in by a time series imputation algorithm, and the timestamp format is unified to obtain a complete and consistent user location record. Based on the user location records and timestamp information, the user's access behavior is extracted, the user's stay records at different points of interest, as well as the arrival and departure times, are identified, and the user's access behavior sequence is generated. Based on the user access behavior sequence and the potential interest region list, a graph matching algorithm is used to transform the user access behavior sequence into a weighted directed graph, where nodes represent interest points and edge weights represent access frequency and dwell time. Through the weighted directed graph, the correlation between the user access behavior sequence and the potential interest region list is established, and the access probability matrix of each interest point is calculated. Using the access probability matrix, a hierarchical density peak clustering algorithm is employed, combined with geofencing technology, to calculate geographical location, access frequency, and dwell time. The points of interest visited by users are grouped, and the clustering parameters are automatically adjusted through a dynamic threshold adjustment mechanism to obtain clustering results that reflect users' frequently visited locations and behavioral patterns. Based on the clustering results, a hidden Markov model is applied to analyze the user's movement state transition patterns in different time periods, identify the user's behavior patterns, and perform semantic clustering on different types of points of interest to identify preferred location types. Based on the user behavior patterns and preferred location types, a comprehensive score is calculated for each point of interest, taking into account geographical distance, access frequency, dwell time, and topic classification, thus constructing a multidimensional user interest model.

2. The method according to claim 1, characterized in that, Based on the user access behavior sequence and the list of potential interest regions, a graph matching algorithm is used to transform the user access behavior sequence into a weighted directed graph, where nodes represent interest points and edge weights represent access frequency and dwell time. Through this weighted directed graph, the correlation between the user access behavior sequence and the list of potential interest regions is established, and the access probability matrix for each interest point is calculated, including: The spatiotemporal consistency check algorithm is used to process the temporal and spatial coherence of the movement path in the user access behavior sequence, resulting in an undirected graph representing the user movement pattern, where nodes represent different points of interest. Based on the user's dwell time and access frequency at each point of interest, and combined with the edge weights optimized by the machine learning model, the undirected graph is weighted and oriented to generate a weighted directed graph that reflects the time sequence of the user's access to points of interest. Based on the weighted directed graph, the user access behavior sequence is compared with the interest point information in the potential interest region list, and a context-aware mechanism is introduced to establish the correlation between the two and obtain the relevance of the adjusted interest points. For each point of interest, based on the relevance of the adjusted point of interest, a Bayesian network is used for probabilistic inference to calculate the probability of the point of interest being visited, forming an access probability matrix.

3. The method according to claim 2, characterized in that, Based on the user's dwell time and access frequency at each point of interest, and combined with edge weights optimized through a machine learning model, the undirected graph is weighted and oriented to generate a weighted directed graph reflecting the temporal order of user access to points of interest, including: The edge weights are optimized using a machine learning model trained on a large amount of historical access data to generate optimized edge weights that can dynamically adapt to changes in user behavior patterns. Based on the optimized edge weights and the timestamp information of user access to points of interest, the directionality of the edges between each pair of points of interest in the undirected graph is determined to obtain a direction definition that reflects the actual access time order of the user. The undirected graph is transformed using the aforementioned directional definition to generate an initial weighted directed graph; Based on anomaly detection algorithms, paths in the weighted directed graph that do not conform to the user's normal behavior patterns are identified and eliminated, resulting in an optimized weighted directed graph that more realistically represents the user's daily movement patterns.

4. The method according to claim 1, characterized in that, Based on the real-time location information, combined with timestamp and velocity vector information, a trajectory prediction algorithm is applied to predict the user's movement trend and generate a list of potential interest regions, including: Using real-time location information processed by the Kalman filter algorithm, combined with timestamp and velocity vector information, the user's historical movement path is reconstructed to obtain a continuous spatiotemporal trajectory. Based on the continuous spatiotemporal trajectory, a multi-step recurrent neural network is used to perform short-term and medium-to-long-term prediction processing on the user's future movement path to obtain multiple predicted movement trend results. By incorporating external environmental data and using a context-aware model to adjust the movement trend results, an optimized movement trend prediction is obtained. Based on the optimized movement trend prediction, and combined with the point of interest database in the geographic information system, the geographical location of the user's arrival is predicted and processed to generate a preliminary list of areas of interest. The spatiotemporal density clustering algorithm is applied to optimize the preliminary list of regions of interest based on the geographical distribution characteristics near the predicted path, resulting in a cluster-optimized list of regions of interest. A personalized preference model is applied to filter the cluster-optimized list of regions of interest, prioritizing those that match the user's preferences to generate a personalized list of potential regions of interest.

5. The method according to claim 1, characterized in that, A dynamic adaptive geofencing system is configured using the user interest model. This system automatically adjusts the geofencing boundaries based on the user's real-time location, behavioral patterns, and changes in the surrounding environment to generate a set of service provision points, including: By using user interest models to comprehensively analyze and process users' real-time location information, user behavior patterns, and preferred location types, the current user's behavioral status and potential destinations can be determined. Based on the user's real-time location information and movement trend results, combined with timestamp information and velocity vector information, a spatiotemporal prediction algorithm is applied to estimate the possible destination locations, resulting in a set of estimated locations. By combining user behavior patterns, preferred location types, and environmental impact factors, the initial boundaries of the geofence are set and processed using machine learning optimization algorithms to obtain the initial geofence boundaries. A dynamic threshold adjustment mechanism is introduced to automatically adjust the initial geofence boundary based on real-time user feedback and new location data, generating an optimized geofence boundary. The application context-aware recommendation system combines users' real-time needs, historical preferences, and the actual availability of service providers to filter and sort the service providers within the optimized geofence, generating an optimized set of service providers.

6. The method according to claim 1, characterized in that, A weighted K-nearest neighbor algorithm is used to sort multiple service providers in the service provider set to obtain a service recommendation list. A multimodal fusion algorithm is then used to combine the geographic location information in the service recommendation list with other sensor data to generate a customized information package, including: Using the weighted K-nearest neighbor algorithm, based on user interest model, real-time location information, and geographical distance, access frequency, dwell time and rating of service providers, multiple service providers in the set of service providers are comprehensively evaluated and ranked to obtain a personalized service recommendation list. A context-aware module is introduced to dynamically adjust the personalized service recommendation list by combining external environment data, thereby generating an optimized service recommendation list; Based on the optimized service recommendation list, a multimodal fusion algorithm is applied to integrate data from different sources, perform multi-dimensional feature description processing on each recommended service provider, and generate service information cards. Natural language processing technology is used to perform semantic understanding and sentiment analysis on the text content in the service information card, extract key information and optimize the expression to obtain the content display format; The content display format is integrated into a comprehensive information structure, and customized information packages are generated based on user behavior patterns.

7. An information push system based on big data matching and GPS positioning, characterized in that, include: The receiving module is used to receive location data updated by high-frequency GPS coordinates from the user's mobile device in real time, and to process the location data using a Kalman filter algorithm to obtain real-time location information. The prediction module is used to predict the user's movement trend based on the real-time location information, combined with timestamp and velocity vector information, and to generate a list of potential interest areas. The identification module is used to analyze the list of potential interest regions and the user's historical location data using a spatial clustering analysis algorithm, identify the user's behavior patterns and preferred location types, and construct a user interest model based on the user's behavior patterns and preferred location types. The configuration module is used to configure a dynamic adaptive geofencing system based on the user interest model. The dynamic adaptive geofencing system is used to automatically adjust the fence boundary according to the user's real-time location, behavior patterns and changes in the surrounding environment to generate a set of service provision points. The generation module is used to sort multiple service providers in the service provider set using a weighted K-nearest neighbor algorithm to obtain a service recommendation list, and to combine the geographical location information in the service recommendation list with other sensor data using a multimodal fusion algorithm to generate a customized information package. Spatial clustering analysis algorithms are used to analyze the list of potential interest regions and the user's historical location data to identify user behavior patterns and preferred location types. Based on these patterns and preferences, a user interest model is constructed, including: The spatiotemporal consistency check algorithm is used to denoise the user's historical location data, removing noise points caused by GPS signal drift. Missing values ​​are filled in by a time series imputation algorithm, and the timestamp format is unified to obtain a complete and consistent user location record. Based on the user location records and timestamp information, the user's access behavior is extracted, the user's stay records at different points of interest, as well as the arrival and departure times, are identified, and the user's access behavior sequence is generated. Based on the user access behavior sequence and the potential interest region list, a graph matching algorithm is used to transform the user access behavior sequence into a weighted directed graph, where nodes represent interest points and edge weights represent access frequency and dwell time. Through the weighted directed graph, the correlation between the user access behavior sequence and the potential interest region list is established, and the access probability matrix of each interest point is calculated. Using the access probability matrix, a hierarchical density peak clustering algorithm is employed, combined with geofencing technology, to calculate geographical location, access frequency, and dwell time. The points of interest visited by users are grouped, and the clustering parameters are automatically adjusted through a dynamic threshold adjustment mechanism to obtain clustering results that reflect users' frequently visited locations and behavioral patterns. Based on the clustering results, a hidden Markov model is applied to analyze the user's movement state transition patterns in different time periods, identify the user's behavior patterns, and perform semantic clustering on different types of points of interest to identify preferred location types. Based on the user behavior patterns and preferred location types, a comprehensive score is calculated for each point of interest, taking into account geographical distance, access frequency, dwell time, and topic classification, thus constructing a multidimensional user interest model.

8. A computing device, characterized in that, It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are invoked and executed by the processing component to implement the information push method based on big data matching and GPS positioning as described in any one of claims 1 to 6.

9. A computer storage medium, characterized in that, The device contains a computer program, which, when executed by a computer, implements an information push method based on big data matching and GPS positioning as described in any one of claims 1 to 6.

Images