Scene service triggering method and device used in multi-fence interaction scene

Abstract

The invention discloses a scene service triggering method and device used in a multi-fence interaction scene, and the method comprises the steps: obtaining the current travel information of a to-be-monitored user ID, determining an electronic fence related to the to-be-monitored user ID based on the travel information, and fitting an overlapping region of electronic fence interaction; when it is monitored that the user enters the overlapping area, historical user travel data corresponding to the ID of the user to be monitored are obtained and input into a pre-trained behavior sequence prediction model, and an action sequence in the multi-fence interaction scene is output; dynamically generating service triggering priorities of the electronic fences of the service providers according to the action sequence; according to the service triggering priorities, the scene services of the electronic fences of the service providers are sorted, and a scene service triggering list is obtained; and triggering the scene service in the list to generate a notification message and feeding the notification message back to the client. Therefore, by adopting the embodiment of the invention, the problem of disordered service sequence in a multi-fence interaction scene can be solved, and the accuracy of service triggering and the service experience degree are improved.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a scene service triggering method and device for use in a multi-fence interaction scene. Background Art

[0002] With the rapid development of the global tourism industry, travelers' demand for hotel, catering and transportation related services is growing. In this context, the related services between hotels, catering and transportation have become particularly important. Travelers expect to enjoy seamless connection and personalized service arrangements between hotels, catering and transportation.

[0003] In related technologies, the services associated with hotels, food orders, and transportation mainly rely on manual or simple automated systems. With the widespread application of smart devices and location services, providing customized services based on users' real-time geographic locations has become a new trend. As one of the applications, electronic fence technology can trigger corresponding services when users enter or leave a specific geographic area.

[0004] However, although the electronic fence technology provides a service triggering mechanism based on geographic location, the existing system has obvious limitations and shortcomings. For example, in cities, due to the density of buildings and areas, different service providers will set up their own electronic fences at the same or similar geographic locations. This will involve multiple fence interactions, which will cause service order confusion. For example, if a user is used to having a meal after getting off the plane, the system will mistakenly judge that the user goes to the hotel first, triggering the reception service. This cannot solve the service order confusion problem in the multi-fence interaction scenario, thereby reducing the accuracy of service triggering and the service experience of passengers. Summary of the invention

[0005] The embodiments of the present application provide a scene service triggering method and device for a multi-fence interaction scene. In order to have a basic understanding of some aspects of the disclosed embodiments, a simple summary is given below. This summary is not a general review, nor is it intended to identify key / important components or describe the scope of protection of these embodiments. Its only purpose is to present some concepts in a simple form as a preface to the detailed description that follows.

[0006] In a first aspect, an embodiment of the present application provides a scene service triggering method for a multi-fence interaction scene, which is applied to a server, and the method includes: Acquire the current travel information of the user ID to be monitored, and determine the electronic fences belonging to each service provider involved in the user ID to be monitored based on the current travel information, and fit the overlapping areas of the electronic fences of each service provider; When it is detected that the ID of the user to be monitored enters the overlapping area, the historical user travel data corresponding to the ID of the user to be monitored is obtained; Input the historical user travel data into the pre-trained behavior sequence prediction model, and output the action sequence corresponding to the monitored user ID in the multi-fence interaction scenario; the pre-trained behavior sequence prediction model is obtained by machine learning based on the historical action sequence labels and historical data of each target user ID in the multi-fence interaction scenario, and the historical data includes historical behavior data, historical context information related to user behavior, and historical user portrait information, and the historical user portrait information is used to characterize the user's basic information and preferences; Dynamically generate the service triggering priority of the geo-fence of each service provider according to the action sequence; According to the service triggering priority of the electronic fence of each service provider, the scene services of the electronic fence of each service provider are sorted to obtain a scene service triggering list; The triggering scene service triggers the scene services in the list to generate notification messages to be fed back to the clients belonging to each service provider.

[0007] Optionally, generate a pre-trained behavior sequence prediction model by following the steps below, including: Collect and preprocess the historical behavior data, historical context information related to user behavior, and historical user profile information of each target user ID in multi-fence interaction scenarios; Obtain the historical action sequence labels of each target user ID in the multi-fence interaction scenario; Create a behavior sequence prediction model, the sequence prediction model includes a decision tree network; Generate model training samples based on historical behavior data, historical context information, and historical user profile information; The model training samples are input into the decision tree network for machine learning, and after the learning is completed, a pre-trained behavior sequence prediction model is obtained.

[0008] Optionally, the behavior sequence prediction model further includes a trend chart fitting module, an interest point area determination module, a preference degree generation module, a coding association module, a feature coding module, a data annotation module and a decision tree network; Generate model training samples based on historical behavior data, historical context information, and historical user profile information, including: The trend chart fitting module fits the factor trend chart of multiple preset selection factors of the target user belonging to each target user ID within a period of time based on the historical behavior data. The preset selection factors are used to describe the basic information of hotels, meal reservations and transportation. The factor trend chart is used to represent the changes in the selection degree of the target user for each preset selection factor at different times; The interest point region determination module determines the region where the selection factor with the most selection times is located from the factor trend graph as the interest point region; The preference generation module generates the preference of each target user for the selection factor according to the interest point area; The coding association module performs coding association processing on the preference degree of each target user for the selection factor and the historical context information of each target user to obtain the first feature; The feature encoding module performs feature encoding on the basic information and preferences of each target user used to characterize the user to obtain a second feature; The data labeling module uses the historical action sequence label of each target user ID to label the first feature and the second feature of each target user to obtain model training samples.

[0009] Optionally, based on historical behavior data, factor trend charts of multiple preset selection factors for target users belonging to each target user ID over a period of time are fitted, including: Determine the time range that needs to be analyzed, and divide the historical behavior data into target behavior data in different time windows based on the time range; Extract features related to preset selection factors from target behavior data; Obtain user selection data for each preset selection factor in different time windows; Perform trend analysis on user selected data to identify trends that change over time and obtain trend analysis results; Using the extracted features and trend analysis results, a trend chart is drawn for each preset selection factor to obtain factor trend charts of multiple preset selection factors for the target user to which each target user ID belongs within a period of time.

[0010] Optionally, based on the interest point area, generate the preference of each target user for the selection factor, including: Aggregate the user's visit frequency, stay duration and consumption records in the POI area to obtain aggregated data; Extract the target user's visit times, average stay time, and consumption amount at the point of interest from the aggregated data to obtain the behavior pattern of each target user at the point of interest; Use clustering algorithms to cluster the behavior patterns of target users at points of interest, so as to classify similar behavior patterns into one category and obtain multiple categories of information; For each type of information, determine the cluster center, which is used to characterize the high-frequency preference of each target user in each type of information; The calculation indicators are set using the activity frequency, activity duration, and consumption amount of each target user in the cluster center; According to the calculation index, the preference intensity of each cluster center is generated as the preference degree of each target user for the selection factor; among them, The calculation formula for the preference strength of each cluster center is:

[0011] in, For the The preference strength of the cluster centers, It is The set of all users in the cluster center, Is a user Activity frequency or consumption amount in the cluster center, Is a user The total dwell time at the cluster center.

[0012] Optionally, the model training samples include key-value pairs of feature factors and historical action sequence labels of each target user ID in a multi-fence interaction scenario, where the feature factors are generated based on historical behavior data, historical context information, and historical user profile information; The model training samples are input into the decision tree network for machine learning. After the learning is completed, the pre-trained behavior sequence prediction model is obtained, including: Input the key-value pairs into the decision tree network for machine learning and output the loss value; When the loss value reaches the minimum, a pre-trained behavior sequence prediction model is generated; or, when the loss value does not reach the minimum, the step of inputting the key-value pair into the decision tree network for machine learning is continued until the loss value reaches the minimum; Among them, the loss function of the decision tree network is:

[0013] in, is the loss value, N is the number of key-value pairs, C is the number of categories of historical action sequence labels, Indicates Is the historical action sequence label of the key-value pair the kind, The decision tree network predicts The historical action sequence label of the key-value pair is The probability of the class.

[0014] Optionally, based on the order of actions, the service triggering priority of the geo-fence of each service provider is dynamically generated, including: Prioritize each action step in the action sequence; Map each action step with a priority number to the electronic fence of each service provider to obtain the initial priority of the electronic fence of each service provider; Obtain context information and real-time location data of the user ID to be detected; According to the context information and real-time location data, the initial priority of the electronic fence of each service provider is adjusted to generate the service triggering priority of the electronic fence of each service provider; wherein the context information includes itinerary, current time, arrival time, weather forecast information, and traffic condition information.

[0015] Optionally, trigger the scene services in the scene service trigger list, including: Get the trigger conditions of each scene service in the scene service trigger list; Starting from the scene service with the highest priority in the scene service trigger list, determine in turn whether the corresponding trigger conditions are met through context information; If yes, trigger the scene service with the highest priority, and continue to execute the step of obtaining the context information and real-time location data of the user ID to be detected, so as to dynamically adjust the initial priority of the electronic fence of each service provider; the method also includes: When it is detected that the user ID to be monitored does not enter the overlapping area and enters the electronic fence of any service provider, the target trigger condition of the entered electronic fence is obtained; When the context information meets the target trigger condition, the target scene service of the entered electronic fence is obtained; Trigger the target scene service.

[0016] Optionally, determining the electronic fences belonging to each service provider involved in the user ID to be monitored based on the current travel information includes: Extract key data from the current information, which is used to characterize each service node in the trip; According to each service node in the itinerary, the corresponding service providers are matched to obtain each service provider involved in the user ID to be monitored; According to each service provider involved in the user ID to be monitored, the corresponding electronic fence is obtained from the pre-established mapping relationship between the service provider and the electronic fence, and the electronic fence belonging to each service provider involved in the user ID to be monitored is obtained.

[0017] In a second aspect, an embodiment of the present application provides a scene service triggering device for a multi-fence interaction scene, the device comprising: An overlapping area fitting module is used to obtain the current travel information of the user ID to be monitored, and determine the electronic fences belonging to each service provider involved in the user ID to be monitored based on the current travel information, and fit the overlapping area of ​​the electronic fence interaction of each service provider; A travel information acquisition module is used to acquire historical user travel data corresponding to the user ID to be monitored when the user ID to be monitored enters the overlapping area; The model processing module is used to input the historical user travel data into the pre-trained behavior sequence prediction model, and output the action sequence corresponding to the monitored user ID in the multi-fence interaction scenario; the pre-trained behavior sequence prediction model is obtained by machine learning based on the historical action sequence label and historical data of each target user ID in the multi-fence interaction scenario, and the historical data includes historical behavior data, historical context information related to user behavior, and historical user portrait information, and the historical user portrait information is used to characterize the basic information and preferences of the user; A service trigger priority generation module is used to dynamically generate the service trigger priority of the electronic fence of each service provider according to the action sequence; A scene service sorting module is used to sort the scene services of the electronic fences of each service provider according to the service triggering priority of the electronic fences of each service provider to obtain a scene service triggering list; The scene service triggering module is used to trigger the scene services in the scene service triggering list to generate notification messages and feed them back to the clients belonging to each service provider.

[0018] The technical solution provided by the embodiments of the present application may have the following beneficial effects: In an embodiment of the present application, by obtaining the user's current travel information and determining the electronic fences of each service provider involved in the user and their overlapping areas, the server can more accurately monitor the user's location changes. After the user is in the overlapping area, by combining historical user travel data and a pre-trained behavior sequence prediction model, the server can accurately predict the user's action sequence and dynamically generate a service trigger priority based on this. At this time, the server can trigger the scene service in the correct order, avoiding the problem of service order disorder, thereby improving the accuracy of service triggering, and service providers can receive service notifications that meet the user's personalization, thereby improving the user experience.

[0019] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0021] Figure 1 It is a method flow diagram of a scene service triggering method for a multi-fence interaction scene provided by an embodiment of the present application; Figure 2 This is a data representation intent of a user's historical behavior data provided by an embodiment of the present application; Figure 3 It is a schematic diagram of the model architecture of a behavior sequence prediction model provided by this application; Figure 4 It is a user selection factor trend chart provided by this application; Figure 5 This is a data representation intent for priority adjustment provided by this application; Figure 6 It is a structural diagram of a scene service triggering device for a multi-fence interaction scene provided by the present application; Figure 7 It is a structural schematic diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0022] The following description and the accompanying drawings sufficiently illustrate specific embodiments of the present application to enable those skilled in the art to practice them.

[0023] It should be clear that the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without creative work are within the scope of protection of the present application.

[0024] When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. Instead, they are only examples of devices and methods consistent with some aspects of the present application as detailed in the attached claims.

[0025] In the description of the present application, it should be understood that the terms "first", "second", etc. are used for descriptive purposes only and should not be understood as indicating or implying relative importance. For those of ordinary skill in the art, the specific meanings of the above terms in the present application can be understood according to the specific circumstances. In addition, in the description of the present application, unless otherwise specified, "multiple" refers to two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the objects associated before and after are in an "or" relationship.

[0026] Currently, hotel, restaurant and transportation services are mainly implemented by manual or simple automated systems. With the widespread application of smart devices and location services, providing customized services based on users' real-time geographic location has become a new trend. As one of the applications, electronic fence technology can trigger corresponding services when users enter or leave a specific geographic area.

[0027] The inventors realize that although the electronic fence technology provides a service triggering mechanism based on geographic location, the existing system has obvious limitations and deficiencies. For example, in a city, due to the density of buildings and areas, different service providers will set up their own electronic fences at the same or similar geographic locations. This will involve multiple fence interactions, which will cause service order confusion. For example, if a user is accustomed to having a meal after getting off the plane, the system will mistakenly judge that the user goes to the hotel first, and the reception service will be triggered at this time. The service order confusion problem in the multi-fence interaction scenario cannot be solved, thereby reducing the accuracy of service triggering and the service experience of passengers.

[0028] In order to solve the above problems, the present application provides a scene service triggering method and device for a multi-fence interaction scenario to solve the problems existing in the above-mentioned related technical issues. In an embodiment of the present application, by obtaining the user's current travel information and determining the electronic fences of each service provider involved in the user and their overlapping areas, the server can more accurately monitor the user's location changes. After the user is in the overlapping area, by combining historical user travel data and a pre-trained behavior sequence prediction model, the server can accurately predict the user's action sequence and dynamically generate a service trigger priority accordingly. At this time, the server can trigger the scene service in the correct order, avoiding the problem of disordered service order, thereby improving the accuracy of service triggering, and the service provider can receive service notifications that meet the user's personalization, which improves the user experience. The following uses an exemplary embodiment for detailed description.

[0029] The following will be combined with the attached Figure 1 -Attached Figure 5 , a scene service triggering method for a multi-fence interaction scenario provided by an embodiment of the present application is introduced in detail. The method can be implemented by a computer program and can be run on a scene service triggering device for a multi-fence interaction scenario based on a von Neumann system. The computer program can be integrated into an application or run as an independent tool application.

[0030] See also Figure 1 , which is a flow chart of a scene service triggering method for a multi-fence interaction scene provided in an embodiment of the present application, and is applied to a server. Figure 1 As shown, the method of the embodiment of the present application may include the following steps: S101, obtaining current travel information of the user ID to be monitored, and determining the electronic fences of each service provider involved in the user ID to be monitored based on the current travel information, and fitting the overlapping areas of the electronic fences of each service provider; Among them, the monitored user ID refers to the unique identifier of a specific user that the service end plans to track and serve. Current travel information refers to the data and information related to the user's current travel, including but not limited to the departure point, destination, departure time, arrival time, mode of transportation, etc. An electronic fence is a virtual geographic boundary, usually defined by GPS or other positioning technologies, which can trigger preset actions or services when the user's device enters or leaves this virtual boundary. Service providers refer to companies or organizations that provide specific services, such as hotels, restaurants, transportation services, etc. Each service provider may set up its own electronic fence to identify their service areas and provide services when users enter these areas. Overlapping areas refer to geographic areas where two or more electronic fences intersect. In a multi-fence interaction scenario, users' movements may cause them to be in the electronic fences of multiple service providers at the same time.

[0031] In some embodiments, for example, passenger A (user ID to be monitored: U12345) plans to travel from home to the airport, then take a flight to another city, and has booked a hotel and dinner at the destination. The server obtains the current travel information of user U12345 from passenger A's mobile application, email communication, and online booking platform, including flight details (departure time, arrival time, airport code), hotel reservation information (hotel location, check-in time), and dinner reservation (restaurant location, reservation time). The server identifies the service providers related to user U12345's itinerary, including airlines, hotels, and restaurants. The server matches the electronic fence of each service provider. For example, the airport fence is set within 1-5 kilometers around the airport, the hotel fence is set within 100-500 meters around the hotel, and the restaurant fence is set within 100-500 meters around the restaurant. Using GIS technology, the server analyzes and determines the overlapping areas of the airport, hotel, and restaurant electronic fences. For example, if a hotel and a restaurant are both located near an airport, their fences may overlap.

[0032] In an embodiment of the present application, the specific process of determining the electronic fences belonging to each service provider involved in the user ID to be monitored based on the current travel information includes: extracting key data from the current appearance information, the key data is used to characterize each service node in the itinerary; matching the corresponding service providers according to each service node in the itinerary, and obtaining the service providers involved in the user ID to be monitored; according to the service providers involved in the user ID to be monitored, obtaining the corresponding electronic fences from the pre-established mapping relationship between the service provider and the electronic fence, and obtaining the electronic fences belonging to each service provider involved in the user ID to be monitored.

[0033] Among them, key data refers to data points extracted from current travel information that are critical to service provision, such as flight number, hotel name, restaurant reservation time, etc. These data are used to determine important nodes in the user's itinerary. Service nodes refer to places or time points in the user's itinerary where specific services are required, such as airport boarding gates, hotel entrances, restaurant tables, etc. The mapping relationship between service providers and electronic fences is a pre-established database or mapping table that records the relationship between each service provider and the electronic fences they set up, which is used for fast search and matching.

[0034] In some embodiments, the server extracts key data from the travel information of user U12345, including: the flight information is flight number XY123, the departure time is 08:00, and the departure place is Capital International Airport. The hotel information is that the hotel name is "Central Hotel" and the estimated arrival time is 14:00. The restaurant reservation is a dinner reservation at "Gourmet Restaurant" and the reservation time is 19:00. The server identifies the service nodes in the user's itinerary, including: the boarding gate of Capital International Airport, the entrance of "Central Hotel", and the table of "Gourmet Restaurant". The server matches the corresponding service provider according to the service node: the airline provides flight services, the "Central Hotel" provides accommodation services, and the "Gourmet Restaurant" provides catering services. The server queries the pre-established mapping relationship between the service provider and the electronic fence, and obtains the following electronic fences: the electronic fence of the boarding gate of Capital International Airport corresponding to the airline, the electronic fence of the hotel entrance corresponding to "Central Hotel", and the electronic fence of the restaurant area corresponding to "Gourmet Restaurant".

[0035] S102, when it is detected that the user ID to be monitored enters the overlapping area, obtaining the historical user travel data corresponding to the user ID to be monitored; Among them, historical user travel data refers to the data records generated by users in past travels, including but not limited to flight records, hotel check-in history, dining consumption habits, transportation mode selection, etc.

[0036] In some embodiments, the user's smartphone or other positioning device sends location updates to the server in real time. The server monitors the user's location through the GIS system to determine when the user enters an overlapping area, for example, when the user is approaching the airport and the booked hotel at the same time. Once the server detects that user U12345 has entered the overlapping area, it retrieves the historical travel data of user U12345 from the database, including: Past flight records: the number of flights that user U12345 has taken in the past year, frequently selected airlines, preferred seat types, etc. Hotel check-in history: user U12345's accommodation preferences, such as room type and whether additional services are needed (such as fitness center, laundry service). Dining consumption habits: user U12345's dining preferences when traveling, such as the types of dishes frequently ordered and favorite restaurant styles. Historical travel data of different users, for example Figure 2 shown.

[0037] S103, inputting historical user travel data into a pre-trained behavior sequence prediction model, and outputting the action sequence corresponding to the monitored user ID in the multi-fence interaction scenario; the pre-trained behavior sequence prediction model is obtained by machine learning based on the historical action sequence labels and historical data of each target user ID in the multi-fence interaction scenario, the historical data including historical behavior data, historical context information related to user behavior, and historical user portrait information, and the historical user portrait information is used to characterize the basic information and preferences of the user; Among them, the pre-trained behavior sequence prediction model is a model pre-trained using machine learning technology. It can predict the user's action sequence in future multi-fence interaction scenarios based on the user's historical travel data. The monitored user ID refers to the unique identifier of a specific user that the server plans to track and serve. The multi-fence interaction scenario refers to the existence of multiple electronic fences in a geographical area. These fences may belong to different service providers, and the user's actions in this area may trigger the services of multiple fences. The action sequence refers to the user's expected action steps and sequence in the multi-fence interaction scenario, such as checking in at the hotel first, then going to the restaurant for dinner, and finally going to the conference center.

[0038] In some embodiments of the present application, the server uses the historical travel data of different users, combined with the travel patterns of other similar users, to train a behavior sequence prediction model. When the new itinerary of user U12345 is confirmed, the server inputs the specific information of this itinerary into the behavior sequence prediction model. The model analyzes the itinerary information and historical preferences of user U12345 to predict his sequence of actions in a multi-fence interaction scenario. The model outputs a prediction result, predicting that user U12345 will act in the following order: arrive at the airport fence (triggering the flight information service), go to the hotel fence (triggering the hotel check-in service), attend the meeting fence (triggering the meeting information service), and finally go to the restaurant fence (triggering the catering service).

[0039] In some embodiments of the present application, a pre-trained behavior sequence prediction model can be generated according to the following steps, specifically including: collecting and pre-processing historical behavior data, historical context information related to user behavior, and historical user portrait information of each target user ID in a multi-fence interaction scenario; obtaining the historical action sequence label of each target user ID in the multi-fence interaction scenario; creating a behavior sequence prediction model, the sequence prediction model includes a decision tree network; based on the historical behavior data, historical context information and historical user portrait information, generating model training samples; inputting the model training samples into the decision tree network for machine learning, and obtaining a pre-trained behavior sequence prediction model after the learning is completed.

[0040] Among them, historical context information refers to environmental or situational information related to user behavior, such as time, weather, traffic conditions, etc., which may affect user behavior. Historical user profile information refers to data describing user characteristics and preferences, such as age, gender, consumption habits, etc., which is used to build user profiles. Historical action sequence labels refer to records of historical action sequences of users in multi-fence interaction scenarios, which are used to train prediction models. Decision tree network is a machine learning model composed of multiple decision trees for classification and regression tasks. Model training samples refer to data sets used to train machine learning models, including input features and corresponding labels.

[0041] In the embodiment of the present application, by collecting and preprocessing the user's historical behavior data, context information and user portrait information, and using these data to train the decision tree network model, a pre-trained behavior sequence prediction model can be created. This model can accurately predict the user's action sequence in a multi-fence interaction scenario, thereby achieving dynamic adjustment and optimization of personalized services, improving the timeliness and relevance of services, and ultimately improving user satisfaction and overall service experience.

[0042] Among them, for example Figure 3 As shown, the behavior sequence prediction model also includes a trend chart fitting module, an interest point area determination module, a preference degree generation module, a coding association module, a feature coding module, a data annotation module and a decision tree network.

[0043] In an embodiment of the present application, the specific process of generating a model training sample based on historical behavior data, historical context information and historical user portrait information includes: a trend chart fitting module fits a factor trend chart of multiple preset selection factors of a target user belonging to each target user ID within a period of time based on historical behavior data, the preset selection factors are used to describe the basic information of hotels, meal reservations and transportation, and the factor trend chart is used to characterize the changes in the degree of selection of each preset selection factor by the target user at different times; the point of interest area determination module determines the area where the selection factor with the most selection times is located from the factor trend chart as the point of interest area; the preference degree generation module generates the preference degree of each target user for the selection factor according to the point of interest area; the encoding association module encodes and associates the preference degree of each target user for the selection factor with the historical context information of each target user to obtain a first feature; the feature encoding module features each target user's basic information and preferences used to characterize the user to obtain a second feature; the data annotation module uses the historical action sequence label of each target user ID to annotate the first feature and the second feature of each target user to obtain a model training sample.

[0044] Among them, the preset selection factors are used to describe the variables of basic information of service providers (such as hotels, restaurants, and transportation), such as the star rating of the hotel, the cuisine of the restaurant, the type of transportation, etc. The factor trend chart describes the changes in the degree of user selection for each preset selection factor at different time points, which is used to visualize the changes in user preferences over time. The point of interest area is the area where the selection factor selected by the user the most times is located, reflecting the user's main interests and preferences. Feature encoding is to convert raw data into a format suitable for machine learning model processing, such as converting categorical data into numerical data.

[0045] In the embodiment of the present application, by analyzing the historical behavior data of each target user and fitting the factor trend chart, the user's interest area and preference level can be identified, and then the feature coding can be generated by combining the user's contextual information and basic information. This process not only enhances the understanding of the user's personalized needs, but also improves the accuracy of the behavior prediction model and the responsiveness of personalized services by creating labeled model training samples, thereby providing users with a more accurate and intimate customized service experience.

[0046] In some embodiments of the present application, based on historical behavior data, the specific process of fitting a factor trend chart of multiple preset selection factors of a target user belonging to each target user ID within a period of time includes: determining the time range to be analyzed, and dividing the historical behavior data into target behavior data of different time windows based on the time range; extracting features related to the preset selection factors from the target behavior data; obtaining user selection data of each preset selection factor in different time windows; performing trend analysis on the user selection data to identify trends that change over time and obtain trend analysis results; using the extracted features and trend analysis results, drawing a trend chart for each preset selection factor, and obtaining a factor trend chart of multiple preset selection factors of the target user belonging to each target user ID within a period of time. Among them, the factor trend chart, for example Figure 4 shown.

[0047] Among them, the time range refers to a span of time determined for a specific analysis purpose, for example, it can be a day, a week, a month or a year. The time window is a smaller time period that further subdivides the time range, which is used to analyze user behavior in more detail. For example, if you analyze user behavior in a day, you can divide the day into three time windows: morning, noon and evening. Target behavior data refers to user behavior data collected within a specific time window. This data is the focus of analysis and is used to extract information related to user behavior patterns and preferences. Feature extraction is to identify and extract information related to preset selection factors from target behavior data. This information can be used to describe user behavior patterns and preferences. User selection data refers to the actual selection records of users for each preset selection factor in different time windows. This data reflects the actual behavior of users. Trend analysis is to analyze user selection data to identify behavior patterns and trends that change over time. The factor trend chart is a chart drawn for each preset selection factor based on the extracted features and trend analysis results, showing the changes in the degree of user selection of these factors over time.

[0048] Specifically, according to the point of interest area, the specific process of generating the preference degree of each target user for the selection factor includes: aggregating the user's visit frequency, stay duration and consumption record included in the behavior data in the point of interest area to obtain aggregated data; extracting the description of the target user's visit times, average stay time and consumption amount at the point of interest from the aggregated data to obtain the behavior pattern of each target user at the point of interest; clustering the behavior patterns of the target users at the point of interest using a clustering algorithm to classify similar behavior patterns into one category to obtain multiple categories of information; for each category of information, determining the cluster center, the cluster center is used to characterize the high-frequency preference of each target user in each category of information; using each target user's activity frequency, activity duration and consumption amount at the cluster center to set calculation indicators; based on the calculation indicators, generating the preference strength of each cluster center as the preference degree of each target user for the selection factor.

[0049] Among them, the calculation formula for the preference strength of each cluster center is:

[0050] in, For the The preference strength of the cluster centers, It is The set of all users in the cluster center, Is a user Activity frequency or consumption amount in the cluster center, Is a user The total dwell time at the cluster center.

[0051] Among them, the model training samples include key-value pairs of feature factors and historical action sequence labels of each target user ID in multi-fence interaction scenarios. The feature factors are generated based on historical behavior data, historical context information and historical user portrait information.

[0052] In some embodiments of the present application, the model training samples are input into a decision tree network for machine learning. After the learning is completed, the specific process of obtaining a pre-trained behavior sequence prediction model includes: inputting key-value pairs into the decision tree network for machine learning, and outputting a loss value; when the loss value reaches the minimum, generating a pre-trained behavior sequence prediction model; or, when the loss value does not reach the minimum, continuing to execute the step of inputting key-value pairs into the decision tree network for machine learning until the loss value reaches the minimum.

[0053] Among them, the loss function of the decision tree network is:

[0054] in, is the loss value, N is the number of key-value pairs, C is the number of categories of historical action sequence labels, Indicates Is the historical action sequence label of the key-value pair the Class (1 for yes, 0 for no), The decision tree network predicts The historical action sequence label of the key-value pair is The probability of the class.

[0055] S104, dynamically generating the service triggering priority of the electronic fence of each service provider according to the action sequence; The service trigger priority refers to the order in which services should be provided or executed first when multiple services may be triggered.

[0056] In some embodiments of the present application, the specific process of dynamically generating the service trigger priority of the electronic fence of each service provider according to the action sequence includes: priority numbering each action step in the action sequence; mapping each action step carrying the priority number to the electronic fence of each service provider to obtain the initial priority of the electronic fence of each service provider; obtaining the context information and real-time location data of the user ID to be detected; adjusting the initial priority of the electronic fence of each service provider according to the context information and the real-time location data to generate the service trigger priority of the electronic fence of each service provider; wherein the context information includes itinerary, current time, arrival time, weather forecast information, and traffic conditions information.

[0057] Among them, the priority number is to assign a number to each action step in the action sequence to indicate their execution priority. An action step refers to a single activity or event in the action sequence, such as "check in to a hotel" or "take a flight". A service provider is a company or organization that provides specific services, such as hotels, catering services, transportation services, etc. The initial priority is the initial priority order assigned to the electronic fences of each service provider based on the priority number in the action sequence. Context information is environmental or situational information related to user behavior, such as itinerary, current time, arrival time, weather forecast information, traffic conditions information, etc.

[0058] For example, user U12345 plans to complete multiple action steps in one day: go to the airport from home (step 1), take a plane to the destination (step 2), go to the hotel from the airport (step 3), and have a meal at a restaurant near the hotel (step 4). The server prioritizes the order of actions according to the itinerary of user U12345, for example: step 1 (go to the airport): priority 1, step 2 (take a plane): priority 2, step 3 (go to the hotel): priority 3, step 4 (meal): priority 4. The server maps each action step to the corresponding service provider's electronic fence and sets the initial priority: airport fence: priority 1, airplane service fence: priority 2, hotel fence: priority 3, restaurant fence: priority 4. The server collects the context information and real-time location data of user U12345, including: itinerary: detailed itinerary of user U12345, current time: departure time is 8 am, arrival time: the plane is expected to arrive at 1 pm, weather forecast information: it will rain in the afternoon at the destination, traffic condition information: traffic congestion at the destination. The server adjusts the service triggering priority of the geo-fences based on the context information and real-time location data: Since it is expected to rain in the afternoon, the priority of the hotel fence is increased so that the user can check in quickly after arriving; considering the traffic congestion, the priority of the restaurant fence is increased so that the dining service can be provided in time after the user arrives at the hotel; according to the real-time location data, when the user U12345 is close to the airport, the priority of the airport fence is increased. The server generates the adjusted service triggering priority of the geo-fences of each service provider: Airport fence (after adjustment): priority 1, hotel fence (after adjustment): priority 2, restaurant fence (after adjustment): priority 3, airplane service fence: priority 4. Among them, the adjusted service triggering priority of the geo-fences of each service provider is, for example Figure 5 shown.

[0059] S105, sorting the scene services of the electronic fences of each service provider according to the service triggering priority of the electronic fences of each service provider to obtain a scene service triggering list; In some embodiments, for example, the server determines the service triggering priority of the electronic fences of each service provider based on the itinerary and context information (such as time, weather, etc.) of the user U12345. For example: airport fence: priority 1 (because the user needs to check in first), hotel fence: priority 2 (the user needs to check in after arriving at the destination), restaurant fence: priority 3 (the user will have a meal after settling in the hotel). At this time, the server sorts the scene services of the electronic fences of each service provider according to the determined priority: scene service trigger list, flight information service triggered by the airport fence, check-in reminder service triggered by the hotel fence, and meal ordering service triggered by the restaurant fence. The server generates the final scene service triggering list: system-triggered flight information service, system-triggered check-in reminder service, and system-triggered meal ordering service.

[0060] S106, triggering the scene services in the scene service triggering list to generate notification messages to be fed back to the clients belonging to each service provider.

[0061] In some embodiments of the present application, the specific process of triggering a scene service in the scene service trigger list includes: obtaining the trigger conditions of each scene service in the scene service trigger list; starting from the scene service with the highest priority in the scene service trigger list, determining in turn through context information whether the corresponding trigger conditions are met; if so, triggering the scene service with the highest priority, and continuing to execute the step of obtaining the context information and real-time location data of the user ID to be detected, so as to dynamically adjust the initial priority of the electronic fence of each service provider.

[0062] In some embodiments of the present application, when it is monitored that the user ID to be monitored has not entered the overlapping area and has entered the electronic fence of any service provider, the target trigger condition of the entered electronic fence is obtained; when the context information meets the target trigger condition, the target scene service of the entered electronic fence is obtained; and the target scene service is triggered.

[0063] In an embodiment of the present application, by obtaining the user's current travel information and determining the electronic fences of each service provider involved in the user and their overlapping areas, the server can more accurately monitor the user's location changes. After the user is in the overlapping area, by combining historical user travel data and a pre-trained behavior sequence prediction model, the server can accurately predict the user's action sequence and dynamically generate a service trigger priority based on this. At this time, the server can trigger the scene service in the correct order, avoiding the problem of service order disorder, thereby improving the accuracy of service triggering, and service providers can receive service notifications that meet the user's personalization, thereby improving the user experience.

[0064] The following is an embodiment of the device of the present application, which can be used to execute the embodiment of the method of the present application. For details not disclosed in the embodiment of the device of the present application, please refer to the embodiment of the method of the present application.

[0065] See also Figure 6 , which shows a schematic diagram of the structure of a scene service triggering device for a multi-fence interaction scenario provided by an exemplary embodiment of the present application. The scene service triggering device for a multi-fence interaction scenario can be implemented as all or part of an electronic device through software, hardware, or a combination of both. The device 1 includes an overlapping area fitting module 10, a travel information acquisition module 20, a model processing module 30, a service trigger priority generation module 40, a scene service sorting module 50, and a scene service triggering module 60.

[0066] The overlapping area fitting module 10 is used to obtain the current travel information of the user ID to be monitored, and determine the electronic fences belonging to each service provider involved in the user ID to be monitored based on the current travel information, and fit the overlapping area of ​​the electronic fence interaction of each service provider; The travel information acquisition module 20 is used to acquire the historical user travel data corresponding to the user ID to be monitored when the user ID to be monitored enters the overlapping area; The model processing module 30 is used to input the historical user travel data into the pre-trained behavior sequence prediction model, and output the action sequence corresponding to the monitored user ID in the multi-fence interaction scenario; the pre-trained behavior sequence prediction model is obtained by machine learning based on the historical action sequence label and historical data of each target user ID in the multi-fence interaction scenario, and the historical data includes historical behavior data, historical context information related to user behavior, and historical user portrait information, and the historical user portrait information is used to characterize the basic information and preferences of the user; A service trigger priority generation module 40 is used to dynamically generate the service trigger priority of the electronic fence of each service provider according to the action sequence; The scene service sorting module 50 is used to sort the scene services of the electronic fences of each service provider according to the service triggering priority of the electronic fences of each service provider to obtain a scene service triggering list; The scene service triggering module 60 is used to trigger the scene services in the scene service triggering list to generate notification messages to be fed back to the clients belonging to each service provider.

[0067] It should be noted that the scene service triggering device for a multi-fence interaction scenario provided in the above embodiment only uses the division of the above functional modules as an example when executing the scene service triggering method for a multi-fence interaction scenario. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the scene service triggering device for a multi-fence interaction scenario provided in the above embodiment and the scene service triggering method embodiment for a multi-fence interaction scenario belong to the same concept, and the implementation process thereof is detailed in the method embodiment, which will not be repeated here.

[0068] The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0069] In an embodiment of the present application, by obtaining the user's current travel information and determining the electronic fences of each service provider involved in the user and their overlapping areas, the server can more accurately monitor the user's location changes. After the user is in the overlapping area, by combining historical user travel data and a pre-trained behavior sequence prediction model, the server can accurately predict the user's action sequence and dynamically generate a service trigger priority based on this. At this time, the server can trigger the scene service in the correct order, avoiding the problem of service order disorder, thereby improving the accuracy of service triggering, and service providers can receive service notifications that meet the user's personalization, thereby improving the user experience.

[0070] The present application also provides a computer-readable medium having program instructions stored thereon, which, when executed by a processor, implement the scene service triggering method for a multi-fence interaction scenario provided by the above-mentioned various method embodiments.

[0071] The present application also provides a computer program product comprising instructions, which, when executed on a computer, enables the computer to execute the scene service triggering method for a multi-fence interaction scenario of each of the above method embodiments.

[0072] See also Figure 7 , is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. Figure 7 As shown, the electronic device 1000 may include: at least one processor 1001 , at least one network interface 1004 , a user interface 1003 , a memory 1005 , and at least one communication bus 1002 .

[0073] The communication bus 1002 is used to realize the connection and communication between these components.

[0074] The user interface 1003 may include a display screen (Display) and a camera (Camera). Optionally, the user interface 1003 may also include a standard wired interface and a wireless interface.

[0075] The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a WI-FI interface).

[0076] Among them, the processor 1001 may include one or more processing cores. The processor 1001 uses various interfaces and lines to connect various parts of the entire electronic device 1000, and executes various functions and processes data of the electronic device 1000 by running or executing instructions, programs, code sets or instruction sets stored in the memory 1005, and calling data stored in the memory 1005. Optionally, the processor 1001 can be implemented in at least one hardware form of digital signal processing (Digital Signal Processing, DSP), field programmable gate array (Field-Programmable Gate Array, FPGA), and programmable logic array (Programmable Logic Array, PLA). The processor 1001 can integrate one or a combination of a central processing unit (Central Processing Unit, CPU), a graphics processing unit (Graphics Processing Unit, GPU) and a modem. Among them, the CPU mainly processes the operating system, user interface and application programs; the GPU is responsible for rendering and drawing the content to be displayed on the display screen; the modem is used to process wireless communications. It can be understood that the above-mentioned modem may not be integrated into the processor 1001, and it can be implemented separately through a chip.

[0077] Among them, the memory 1005 may include a random access memory (Random Access Memory, RAM) and may also include a read-only memory (Read-Only Memory). Optionally, the memory 1005 includes a non-transitory computer-readable storage medium. The memory 1005 can be used to store instructions, programs, codes, code sets or instruction sets. The memory 1005 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as a touch function, a sound playback function, an image playback function, etc.), instructions for implementing the above-mentioned various method embodiments, etc.; the data storage area may store data involved in the above-mentioned various method embodiments, etc. The memory 1005 may also be optionally at least one storage system located away from the aforementioned processor 1001. As Figure 7 As shown, the memory 1005 as a computer storage medium may include an operating system, a network communication module, a user interface module, and a scene service triggering application for a multi-fence interaction scene.

[0078] exist Figure 7 In the electronic device 1000 shown, the user interface 1003 is mainly used to provide an input interface for the user and obtain data input by the user; and the processor 1001 can be used to call the scene service trigger application for the multi-fence interaction scene stored in the memory 1005, and specifically perform the following operations: Acquire the current travel information of the user ID to be monitored, and determine the electronic fences belonging to each service provider involved in the user ID to be monitored based on the current travel information, and fit the overlapping areas of the electronic fences of each service provider; When it is detected that the ID of the user to be monitored enters the overlapping area, the historical user travel data corresponding to the ID of the user to be monitored is obtained; Input the historical user travel data into the pre-trained behavior sequence prediction model, and output the action sequence corresponding to the monitored user ID in the multi-fence interaction scenario; the pre-trained behavior sequence prediction model is obtained by machine learning based on the historical action sequence labels and historical data of each target user ID in the multi-fence interaction scenario, and the historical data includes historical behavior data, historical context information related to user behavior, and historical user portrait information, and the historical user portrait information is used to characterize the user's basic information and preferences; Dynamically generate the service triggering priority of the geo-fence of each service provider according to the action sequence; According to the service triggering priority of the electronic fence of each service provider, the scene services of the electronic fence of each service provider are sorted to obtain a scene service triggering list; The triggering scene service triggers the scene services in the list to generate notification messages to be fed back to the clients belonging to each service provider.

[0079] In one embodiment, when the processor 1001 generates a pre-trained behavior sequence prediction model, the processor 1001 specifically performs the following operations: Collect and preprocess the historical behavior data, historical context information related to user behavior, and historical user profile information of each target user ID in multi-fence interaction scenarios; Obtain the historical action sequence labels of each target user ID in the multi-fence interaction scenario; Create a behavior sequence prediction model, the sequence prediction model includes a decision tree network; Generate model training samples based on historical behavior data, historical context information, and historical user profile information; The model training samples are input into the decision tree network for machine learning, and after the learning is completed, a pre-trained behavior sequence prediction model is obtained.

[0080] In one embodiment, when the processor 1001 generates a model training sample based on historical behavior data, historical context information, and historical user portrait information, the processor 1001 specifically performs the following operations: The trend chart fitting module fits the factor trend chart of multiple preset selection factors of the target user belonging to each target user ID within a period of time based on the historical behavior data. The preset selection factors are used to describe the basic information of hotels, meal reservations and transportation. The factor trend chart is used to represent the changes in the selection degree of the target user for each preset selection factor at different times; The interest point region determination module determines the region where the selection factor with the most selection times is located from the factor trend graph as the interest point region; The preference generation module generates the preference of each target user for the selection factor according to the interest point area; The coding association module performs coding association processing on the preference degree of each target user for the selection factor and the historical context information of each target user to obtain the first feature; The feature encoding module performs feature encoding on the basic information and preferences of each target user used to characterize the user to obtain a second feature; The data labeling module uses the historical action sequence label of each target user ID to label the first feature and the second feature of each target user to obtain model training samples.

[0081] In one embodiment, when the processor 1001 performs fitting of a factor trend chart of multiple preset selection factors of a target user to which each target user ID belongs within a period of time based on historical behavior data, the processor 1001 specifically performs the following operations: Determine the time range that needs to be analyzed, and divide the historical behavior data into target behavior data in different time windows based on the time range; Extract features related to preset selection factors from target behavior data; Obtain user selection data for each preset selection factor in different time windows; Perform trend analysis on user selected data to identify trends that change over time and obtain trend analysis results; Using the extracted features and trend analysis results, a trend chart is drawn for each preset selection factor to obtain factor trend charts of multiple preset selection factors for the target user to which each target user ID belongs within a period of time.

[0082] In one embodiment, when the processor 1001 generates the preference degree of each target user for the selection factor according to the interest point area, the processor 1001 specifically performs the following operations: Aggregate the user's visit frequency, stay duration and consumption records in the POI area to obtain aggregated data; Extract the target user's visit times, average stay time, and consumption amount at the point of interest from the aggregated data to obtain the behavior pattern of each target user at the point of interest; Use clustering algorithms to cluster the behavior patterns of target users at points of interest, so as to classify similar behavior patterns into one category and obtain multiple categories of information; For each type of information, determine the cluster center, which is used to characterize the high-frequency preference of each target user in each type of information; The calculation indicators are set using the activity frequency, activity duration, and consumption amount of each target user in the cluster center; According to the calculation index, the preference intensity of each cluster center is generated as the preference degree of each target user for the selection factor; among them, The calculation formula for the preference strength of each cluster center is:

[0083] in, For the The preference strength of the cluster centers, It is The set of all users in the cluster center, Is a user Activity frequency or consumption amount in the cluster center, Is a user The total dwell time at the cluster center.

[0084] In one embodiment, when the processor 1001 inputs the model training sample into the decision tree network for machine learning and obtains the pre-trained behavior sequence prediction model after the learning is completed, the processor 1001 specifically performs the following operations: Input the key-value pairs into the decision tree network for machine learning and output the loss value; When the loss value reaches the minimum, a pre-trained behavior sequence prediction model is generated; or, when the loss value does not reach the minimum, the step of inputting the key-value pair into the decision tree network for machine learning is continued until the loss value reaches the minimum; Among them, the loss function of the decision tree network is:

[0085] in, is the loss value, N is the number of key-value pairs, C is the number of categories of historical action sequence labels, Indicates Is the historical action sequence label of the key-value pair the Class (1 for yes, 0 for no), The decision tree network predicts The historical action sequence label of the key-value pair is The probability of the class.

[0086] In one embodiment, when the processor 1001 dynamically generates the service triggering priority of the electronic fence of each service provider according to the action sequence, the processor 1001 specifically performs the following operations: Prioritize each action step in the action sequence; Map each action step with a priority number to the electronic fence of each service provider to obtain the initial priority of the electronic fence of each service provider; Obtain context information and real-time location data of the user ID to be detected; According to the context information and real-time location data, the initial priority of the electronic fence of each service provider is adjusted to generate the service triggering priority of the electronic fence of each service provider; wherein the context information includes itinerary, current time, arrival time, weather forecast information, and traffic condition information.

[0087] In one embodiment, when executing the scene service in the triggering scene service triggering list, the processor 1001 specifically performs the following operations: Get the trigger conditions of each scene service in the scene service trigger list; Starting from the scene service with the highest priority in the scene service trigger list, determine in turn whether the corresponding trigger conditions are met through context information; If yes, trigger the scene service with the highest priority, and continue to execute the step of obtaining the context information and real-time location data of the user ID to be detected, so as to dynamically adjust the initial priority of the electronic fence of each service provider; the method also includes: When it is detected that the user ID to be monitored does not enter the overlapping area and enters the electronic fence of any service provider, the target trigger condition of the entered electronic fence is obtained; When the context information meets the target trigger condition, the target scene service of the entered electronic fence is obtained; Trigger the target scene service.

[0088] In one embodiment, when the processor 1001 determines the electronic fences belonging to each service provider and related to the user ID to be monitored based on the current travel information, the processor 1001 specifically performs the following operations: Extract key data from the current information, which is used to characterize each service node in the trip; According to each service node in the itinerary, the corresponding service providers are matched to obtain each service provider involved in the user ID to be monitored; According to each service provider involved in the user ID to be monitored, the corresponding electronic fence is obtained from the pre-established mapping relationship between the service provider and the electronic fence, and the electronic fence belonging to each service provider involved in the user ID to be monitored is obtained.

[0089] In an embodiment of the present application, by obtaining the user's current travel information and determining the electronic fences of each service provider involved in the user and their overlapping areas, the server can more accurately monitor the user's location changes. After the user is in the overlapping area, by combining historical user travel data and a pre-trained behavior sequence prediction model, the server can accurately predict the user's action sequence and dynamically generate a service trigger priority based on this. At this time, the server can trigger the scene service in the correct order, avoiding the problem of service order disorder, thereby improving the accuracy of service triggering, and service providers can receive service notifications that meet the user's personalization, thereby improving the user experience.

[0090] A person skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the program for triggering the scene service in the multi-fence interaction scenario can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the storage medium of the program for triggering the scene service in the multi-fence interaction scenario can be a disk, an optical disk, a read-only storage memory, or a random access memory, etc.

[0091] The above disclosure is only the preferred embodiment of the present application, which certainly cannot be used to limit the scope of rights of the present application. Therefore, equivalent changes made according to the claims of the present application are still within the scope covered by the present application.

Claims

1. A scene service triggering method for a multi-fence interaction scene, characterized in that: Applied to the server, the method includes: Acquire current travel information of the user ID to be monitored, and determine the electronic fences of each service provider involved in the user ID to be monitored based on the current travel information, and fit the overlapping areas of the electronic fences of each service provider; When it is detected that the user ID to be monitored enters the overlapping area, obtaining historical user travel data corresponding to the user ID to be monitored; Input the historical user travel data into a pre-trained behavior sequence prediction model, and output the action sequence corresponding to the monitored user ID in the multi-fence interaction scenario; the pre-trained behavior sequence prediction model is obtained by machine learning based on the historical action sequence labels and historical data of each target user ID in the multi-fence interaction scenario, the historical data includes historical behavior data, historical context information related to user behavior, and historical user portrait information, and the historical user portrait information is used to characterize the user's basic information and preferences; Dynamically generate the service triggering priority of the electronic fence of each service provider according to the action sequence; According to the service triggering priority of the electronic fence of each service provider, the scene services of the electronic fence of each service provider are sorted to obtain a scene service triggering list; The scene services in the scene service triggering list are triggered to generate notification messages to be fed back to the clients belonging to the service providers.

2. The method according to claim 1, characterized in that The following steps are used to generate a pre-trained behavior sequence prediction model, including: Collect and preprocess the historical behavior data, historical context information related to user behavior, and historical user profile information of each target user ID in multi-fence interaction scenarios; Obtain the historical action sequence labels of each target user ID in the multi-fence interaction scenario; Creating a behavior sequence prediction model, wherein the sequence prediction model includes a decision tree network; Generate a model training sample based on the historical behavior data, the historical context information, and the historical user portrait information; The model training samples are input into the decision tree network for machine learning, and a pre-trained behavior sequence prediction model is obtained after the learning is completed.

3. The method according to claim 2, characterized in that The behavior sequence prediction model also includes a trend chart fitting module, an interest point area determination module, a preference degree generation module, a coding association module, a feature coding module, a data annotation module and a decision tree network; The generating a model training sample based on the historical behavior data, the historical context information and the historical user portrait information includes: The trend chart fitting module fits the factor trend chart of multiple preset selection factors of the target user to which each target user ID belongs within a period of time based on the historical behavior data, wherein the preset selection factors are used to describe basic information of hotels, meal reservations and transportation tools, and the factor trend chart is used to represent the change in the selection degree of the target user for each preset selection factor at different times; The interest point region determination module determines, from the factor trend graph, a region where the selection factor with the most selection times is located as the interest point region; The preference level generating module generates the preference level of each target user for the selection factor according to the interest point area; The coding association module performs coding association processing on the preference degree of each target user for the selection factor and the historical context information of each target user to obtain a first feature; The feature encoding module performs feature encoding on the basic information and preferences of each target user used to characterize the user to obtain a second feature; The data labeling module uses the historical action sequence label of each target user ID to label the first feature and the second feature of each target user to obtain a model training sample.

4. The method according to claim 3, characterized in that The factor trend chart of fitting multiple preset selection factors of the target user to which each target user ID belongs within a period of time based on the historical behavior data includes: Determine a time range to be analyzed, and divide the historical behavior data into target behavior data of different time windows based on the time range; Extracting features related to preset selection factors from the target behavior data; Obtaining user selection data for each preset selection factor within the different time windows; Performing trend analysis on the user selection data to identify trends that change over time and obtain trend analysis results; Using the extracted features and the trend analysis results, a trend chart is drawn for each preset selection factor to obtain factor trend charts of multiple preset selection factors of the target user to which each target user ID belongs within a period of time.

5. The method according to claim 3, characterized in that: Generating the preference degree of each target user for the selection factor according to the interest point area includes: Aggregate the user's visit frequency, stay duration and consumption record included in the behavior data in the POI area to obtain aggregated data; Extracting the visit times, average stay time and consumption amount of the target user at the point of interest from the aggregated data to obtain the behavior pattern of each target user at the point of interest; Clustering the behavior patterns of the target users at the points of interest using a clustering algorithm to classify similar behavior patterns into one category, thereby obtaining multiple categories of information; For each type of information, determine a cluster center, where the cluster center is used to characterize the high-frequency preference of each target user in each type of information; The calculation index is set by using the activity frequency, activity duration and consumption amount of each target user in the cluster center; According to the calculation index, the preference strength of each cluster center is generated as the preference degree of each target user for the selection factor; wherein, The calculation formula for the preference strength of each cluster center is: in, For the The preference strength of the cluster centers, It is The set of all users in the cluster center, Is a user Activity frequency or consumption amount in the cluster center, Is a user The total dwell time at the cluster center.

6. The method according to claim 2, characterized in that The model training samples include a key-value pair of a feature factor and a historical action sequence label of each target user ID in a multi-fence interaction scenario, wherein the feature factor is generated based on the historical behavior data, the historical context information, and the historical user portrait information; The step of inputting the model training samples into the decision tree network for machine learning, and obtaining a pre-trained behavior sequence prediction model after the learning is completed, comprises: Input the key-value pair into the decision tree network for machine learning, and output a loss value; When the loss value reaches the minimum, generating a pre-trained behavior sequence prediction model; or, when the loss value does not reach the minimum, continuing to perform the step of inputting the key-value pair into the decision tree network for machine learning until the loss value reaches the minimum; Among them, the loss function of the decision tree network is: in, is the loss value, N is the number of key-value pairs, C is the number of categories of historical action sequence labels, Indicates Is the historical action sequence label of the key-value pair the kind, The decision tree network predicts The historical action sequence label of the key-value pair is The probability of the class.

7. The method according to claim 1, characterized in that The dynamically generating the service triggering priority of the electronic fence of each service provider according to the action sequence includes: Prioritize each action step in the action sequence; Mapping each action step with a priority number to the electronic fence of each service provider to obtain the initial priority of the electronic fence of each service provider; Obtaining context information and real-time location data of the user ID to be detected; According to the context information and real-time location data, the initial priority of the electronic fence of each service provider is adjusted to generate the service triggering priority of the electronic fence of each service provider; wherein the context information includes itinerary, current time, arrival time, weather forecast information, and traffic condition information.

8. The method according to claim 7, characterized in that Triggering the scene service in the scene service triggering list includes: Obtaining the triggering condition of each scene service in the scene service triggering list; Starting from the scene service with the highest priority in the scene service trigger list, determining in turn through the context information whether the trigger conditions corresponding thereto are met; If yes, trigger the scene service with the highest priority, and continue to execute the step of obtaining the context information and real-time location data of the user ID to be detected, so as to dynamically adjust the initial priority of the electronic fence of each service provider; the method also includes: When it is detected that the user ID to be monitored does not enter the overlapping area but enters the electronic fence of any service provider, obtaining the target trigger condition of the entered electronic fence; When the context information satisfies the target trigger condition, obtaining a target scene service of the entered electronic fence; The target scenario service is triggered.

9. The method according to claim 1, characterized in that: The step of determining the electronic fences belonging to each service provider and involved in the user ID to be monitored based on the current travel information includes: Extracting key data from the current occurrence information, wherein the key data is used to characterize each service node in the trip; According to each service node in the itinerary, the corresponding service provider is matched to obtain each service provider involved in the user ID to be monitored; According to each service provider involved in the user ID to be monitored, the corresponding electronic fence is obtained from the pre-established mapping relationship between the service provider and the electronic fence, and the electronic fence belonging to each service provider involved in the user ID to be monitored is obtained.

10. A scene service triggering device for a multi-fence interaction scene, characterized in that: The device comprises: An overlapping area fitting module, used to obtain the current travel information of the user ID to be monitored, and determine the electronic fences belonging to each service provider involved in the user ID to be monitored based on the current travel information, and fit the overlapping area of ​​the electronic fences of each service provider; A travel information acquisition module, used for acquiring historical user travel data corresponding to the user ID to be monitored when the user ID to be monitored enters the overlapping area; A model processing module is used to input the historical user travel data into a pre-trained behavior sequence prediction model, and output the action sequence corresponding to the monitored user ID in the multi-fence interaction scenario; the pre-trained behavior sequence prediction model is obtained by machine learning based on the historical action sequence label and historical data of each target user ID in the multi-fence interaction scenario, the historical data includes historical behavior data, historical context information related to user behavior, and historical user portrait information, and the historical user portrait information is used to characterize the user's basic information and preferences; A service trigger priority generation module, used to dynamically generate the service trigger priority of the electronic fence of each service provider according to the action sequence; A scene service sorting module, used to sort the scene services of the electronic fences of the service providers according to the service triggering priority of the electronic fences of the service providers, and obtain a scene service triggering list; The scene service triggering module is used to trigger the scene services in the scene service triggering list to generate notification messages and feed them back to the clients belonging to the service providers.

Images