Information push method and device
By predicting the time when the terminal device reaches the target area and combining with user travel habits, the problem of unreasonable service information push timing under community positioning technology is solved, more accurate service information push is achieved, and user experience and accuracy are improved.
Patent Information
- Application Number
- CN202410381498.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-11-24
- Publication Date
- 2024-11-05
- Estimated Expiration
- 2043-11-24
AI Technical Summary
When terminal devices push service information through cell positioning technology, there is a problem of error triggering, resulting in unreasonable timing of service information push and affecting user experience.
By collecting cell information accessed by terminal devices, predicting the time when the user arrives at the target area, pushing service information for target behavior, so as to improve the accuracy of arrival time prediction, and combining the user's travel habits and time rules, the time prediction model is trained to Improve accuracy.
Push service information at the right time to improve user experience, reduce power consumption, and enhance the accuracy of service information.
Smart Images

Figure CN118200845B_ABST
Abstract
Description
[0001] This application is a divisional application. The application number of the original application is 202311576011.X, and the original application date is November 24, 2023. The entire contents of the original application are incorporated into this application by reference. Technical Field
[0002] The present application relates to the field of terminal positioning technology, and specifically to an information push method and device. Background Art
[0003] The terminal device can determine its own location through positioning technology. In some application scenarios, when the terminal device arrives at a certain location, the relevant application or service can push location-related service information to the terminal device.
[0004] In specific implementation, the terminal device can determine whether it has reached the corresponding location through the information of the connected cell (e.g., the cell identifier (ID)). That is, when the terminal device is connected to the target cell, it is considered that the terminal device has reached the corresponding location, and the service information can be triggered. However, the coverage of some cells is large, and the actual location of the terminal device after connecting to the target cell may be far away from the corresponding location, resulting in false triggering of service information and affecting user experience. Summary of the invention
[0005] In view of the problem in the prior art that a terminal device cannot accurately determine whether it has reached a corresponding location, resulting in false triggering of service information, the present application provides an information push method and device.
[0006] In a first aspect, an embodiment of the present invention provides an information push method, the method comprising:
[0007] Continuously collect cell information accessed by the terminal device at preset time intervals;
[0008] When the cell accessed by the terminal device is switched from the first cell to the second cell, and the second cell is the target cell, predicting the first time according to the continuously collected cell information;
[0009] Push service information for target behavior as soon as possible.
[0010] The embodiment of the present invention collects the cells connected to the terminal device, determines the route of the terminal device to the target area, and predicts the time required for the user to enter the target cell and actually reach the target area by learning the travel habits and time patterns on the corresponding route, so as to improve the accuracy of the arrival time prediction, ensure that the service information is pushed at the right time, and take into account low power consumption and high accuracy at the same time, thereby improving the user experience.
[0011] Optionally, in order to improve the accuracy of the predicted arrival time, it is necessary to determine the timing of executing the time prediction. Considering that the distance from the access target cell to the arrival target area is relatively fixed and is less likely to be interfered by other factors, the time when the second cell is determined to be the target cell is used as the timing of executing the time prediction, wherein the second cell is the target cell, including:
[0012] The second cell is a cell in the first cell list;
[0013] The first cell list is used to record cells covering the geographical location where the user is located when the target behavior successfully occurs, and the target behavior is a user operation instruction corresponding to the service information.
[0014] Optionally, in order to improve the accuracy of the predicted arrival time, considering that different travel habits of users may affect the arrival time, in order to eliminate the influence, it is necessary to comprehensively consider the travel information of the user, and the above method further includes:
[0015] Collect at least one of the following information:
[0016] Transportation information, travel date information, and travel time period information under the current travel date;
[0017] The collected information is used together with the cell information to predict the first time.
[0018] Optionally, the first time is predicted by training a time prediction model. At the same time, in order to learn the user's travel habits to the target area via different routes, respective time prediction models are trained for different routes. During actual travel, the most matching time prediction model is selected to perform time prediction to improve the accuracy of the predicted arrival time. The first time is predicted based on the continuously collected cell information, including:
[0019] Inputting the continuously collected cell information into a first time prediction model, the first time prediction model is used to predict the duration according to the continuously collected cell information, and taking the duration output by the first time prediction model as the first time;
[0020] The first time prediction model is the time prediction model corresponding to the second cell list to which the second cell belongs. The cells recorded in the second cell list are cells whose geographical locations are located when the distance between the geographical locations where the user successfully performs the target behavior is less than the first distance threshold. Several second cell lists constitute the first cell list.
[0021] Optionally, in order to ensure that the result of the model prediction does not have too large an error, before the continuously collected cell information is input into the first time prediction model, the method further includes:
[0022] It is determined that the model error of the first time prediction model is less than a preset threshold.
[0023] Optionally, when the model has a large error, in order to prevent erroneous prediction information from reducing user experience, the time prediction model is no longer used to predict the arrival time, which also includes:
[0024] When it is determined that the model error of the first-time prediction model is not less than a preset threshold, the service information is pushed.
[0025] Optionally, in order to facilitate the training of the time prediction model, the continuously collected cell information is input into the first time prediction model, including:
[0026] According to the order of collection time, the continuously collected cell information is encoded into a route feature sequence;
[0027] The route feature sequence is input into the first time prediction model.
[0028] Optionally, in order to perform training of the time prediction model, it is necessary to collect historical travel data and learn the travel habits of users on each route based on the historical travel data. The method also includes:
[0029] Obtain the first historical travel data where the target behavior occurs;
[0030] A location cluster is obtained according to the first historical travel data clustering;
[0031] A second cell list corresponding to each location cluster is determined according to the first historical travel data corresponding to each location cluster, and a time prediction model corresponding to each location cluster is trained.
[0032] Optional, first historical travel data, including:
[0033] The geographical location of the terminal device when the target behavior occurs, the third cell connected when the target behavior occurs, the first time interval between the time when the target behavior occurs and the first connection to the third cell, the cell information to which the terminal device has been connected in sequence before the target behavior occurs, and the cell information of other cells that the terminal device can access before the target behavior occurs.
[0034] Optionally, in order to determine the common routes of users to the target area for distinction, several geographical locations where the target behavior occurs and are close to each other can be considered to be reached by the same route, and location clusters are obtained according to the first historical travel data, including:
[0035] According to the distances between the geographical locations where the target behavior occurred in the first historical travel data, the geographical locations whose distances are less than a first distance threshold are clustered to obtain a plurality of location clusters.
[0036] Optionally, when the model training is specifically performed, determining the second cell list corresponding to each location cluster according to the first historical travel data corresponding to each location cluster and training the time prediction model corresponding to each location cluster includes:
[0037] Encode the third cell in the first historical travel data contained in each location cluster, the cell information that the terminal device has connected to in sequence before the target behavior occurs, and the cell information of other cells that the terminal device can access before the target behavior occurs to obtain feature data;
[0038] The first time interval in the first historical travel data contained in each location cluster is used as a training label;
[0039] Dividing the first historical travel data including feature data and training labels into training samples and evaluation samples;
[0040] The time prediction model corresponding to each location cluster is obtained based on the training samples and evaluation samples.
[0041] Optionally, after completing the model training, the time prediction model may be updated according to the subsequently collected travel data to further improve the prediction accuracy of the time prediction model for the first time, wherein after pushing the service information for the target behavior after the first time, the method further includes:
[0042] When the target behavior is detected, the cell information continuously collected by the terminal device, the geographical location of the terminal device when the target behavior occurs, the third cell connected when the target behavior occurs, the first time interval between the time when the target behavior occurs and the first connection to the third cell, and the cell information of other cells accessible to the terminal device before the target behavior occurs are determined as the second historical travel data;
[0043] The existing time prediction model is retrained at a preset time according to the second historical travel data.
[0044] Optionally, when updating the training time prediction model, the newly collected historical travel data may include a new route to the target area, so it is necessary to simultaneously update the location clusters and the time prediction models corresponding to each location cluster, and retrain the existing time prediction model based on the second historical travel data, including:
[0045] According to the geographical location where the terminal device is located when the target behavior occurs in the second historical travel data, geographical locations whose distances from other geographical locations where the terminal device is located when the target behavior occurs are less than a first distance threshold are allocated to the existing location cluster;
[0046] Divide new location clusters according to other geographical locations where the terminal device is located when the target behavior occurs in the second historical travel data;
[0047] The second cell lists corresponding to each location cluster are updated, and the time prediction models corresponding to each second cell list are updated and trained according to the second historical travel data.
[0048] Optionally, after completing the update of the time prediction model, it is necessary to ensure that the error of the model is reduced and the accuracy is improved. The method further includes:
[0049] If the model error of the retrained time prediction model is less than a preset threshold, the time prediction model is updated;
[0050] If the model error of the retrained time prediction model is greater than or equal to a preset threshold, the time prediction model is kept unchanged.
[0051] Optionally, push service information after the first time, including:
[0052] The service information is displayed on the display interface of the terminal device.
[0053] In a second aspect, an embodiment of the present invention provides an electronic device, comprising a memory for storing computer program instructions and a processor for executing the computer program instructions, wherein when the computer program instructions are executed by the processor, the electronic device is triggered to execute any method step as in the first aspect.
[0054] In a third aspect, an embodiment of the present invention provides a computer-readable storage medium, in which a computer program is stored, and when the computer-readable storage medium is run on a computer, the computer executes any method in the first aspect. BRIEF DESCRIPTION OF THE DRAWINGS
[0055] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0056] like Figure 1A FIG. 1 is a schematic diagram of an embodiment of information push provided by the present invention;
[0057] like Figure 1B FIG. 1 is a schematic diagram of an embodiment of information push provided by the present invention;
[0058] like Figure 2 FIG. 1 is a schematic diagram of an embodiment of information push provided by the present invention;
[0059] like Figure 3 FIG. 1 is a flow chart of a data collection method provided by an embodiment of the present invention;
[0060] like Figure 4 , which is a flow chart of an information push method provided by an embodiment of the present invention;
[0061] like Figure 5 FIG. 1 is a schematic diagram of a specific embodiment of a clustering position cluster provided by an embodiment of the present invention;
[0062] like Figure 6 , which is a schematic diagram of a location cluster provided by an embodiment of the present invention;
[0063] like Figure 7 , which is a flow chart of a model training method provided by an embodiment of the present invention;
[0064] like Figure 8 , which is a flow chart of an information push method provided by an embodiment of the present invention;
[0065] like Fig. 9 , which is a flow chart of an information push method provided by an embodiment of the present invention;
[0066] like Fig.10 , which is a schematic diagram of the structure of a terminal device provided by the present invention;
[0067] like Fig.11 FIG. 1 is a flowchart of a specific embodiment of the information push method provided by the present invention;
[0068] like Fig.12 , which is a schematic diagram of the hardware structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION
[0069] In order to better understand the technical solution of the present application, the embodiments of the present application are described in detail below with reference to the accompanying drawings.
[0070] It should be clear that the described embodiments are only part of the embodiments of the present application, not all of 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.
[0071] Before describing the embodiments of the present application, the related technologies and technical problems are first described.
[0072] Terminal devices generally have positioning functions. Terminal devices can use positioning technology to determine their real-time location, thereby determining the location information of the user holding the terminal device and providing various services to the user holding the terminal device. For example, when the terminal device arrives at a certain target area, the terminal device can push service information related to the target area through related applications or services for the user to operate, thereby reducing the user's operation steps, or reminding the user when the user forgets to perform related operations.
[0073] Among them, the positioning function of the terminal device generally includes GPS positioning, cell positioning, etc. GPS positioning has high accuracy, but frequent use of GPS positioning will cause high power consumption of the terminal device. In order to avoid excessive power consumption loss of the terminal device, the location information of the terminal device is generally determined through cell positioning.
[0074] like Figure 1A and Figure 1B As shown, taking the example of pushing punch-in service information to users based on cell positioning technology, it is assumed that the circular area with a radius of 500 meters with the work unit as the center is the punch-in area. After the user enters the punch-in area with the terminal device, he can successfully punch in on the terminal device. In order to facilitate user operation, the terminal device provides the user with a punch-in service. After the user subscribes to the punch-in service, the terminal device determines whether the terminal device enters the punch-in area through the cell positioning technology, and pushes the punch-in service information after entering the punch-in area, prompting the user to execute the punch-in. Specifically, the terminal device records the target cell, which is the cell to which the terminal device is connected when the historical punch-in is successful. When the terminal device is connected to the target cell, the terminal device determines that it has entered the punch-in area and pushes the punch-in service information to the user. The user can link to the punch-in application through the punch-in service information and execute the punch-in in the punch-in application. For example, the terminal device records the cell 1 to which the user successfully punches in, and when the terminal device is connected to the cell 1, the punch-in service information is pushed to the user.
[0075] However, the above method of pushing check-in service information to users based on cell positioning technology has the problems of unreasonable service information push timing and erroneous push.
[0076] After analysis, it was found that due to the large coverage area of the cell, the cell coverage area is not necessarily completely located in the punch-in area. The method of using the cell to which the terminal device was connected when the punch-in was successful in the past to determine whether it has entered the punch-in area is not accurate. Figure 1A For example, if the user successfully clocks in when connected to cell 1, the terminal device will record cell 1 as the target cell. However, due to the large coverage of cell 1, only part of cell 1 is located in the clock-in area, which overlaps with the clock-in area. As a result, the terminal device may not actually enter the clock-in area when accessing cell 1. Figure 1AAs shown, assuming that a user holds a terminal device and enters the coverage of cell 1 from position B and accesses cell 1, the terminal device will determine that the user has entered the check-in area and push the check-in service information. After observing the check-in service information pushed by the terminal device, the user can click to enter the check-in interface and execute the check-in in the check-in interface. However, since the terminal device is located at position B at this time and has not entered the check-in area, the user cannot check in successfully, and the check-in application will display the corresponding prompt information of the failed check-in. Since the terminal device has pushed the check-in service information to the user, when the user actually enters the check-in area, such as position A, the terminal device will not push the check-in service on the display interface again, and the user still needs to open the check-in application by himself to complete the check-in.
[0077] It can be seen that due to the large coverage of the cell, the coverage of the cell to which the terminal device is connected when the user successfully clocks in is not completely within the clock-in area, and the location of the cell to which the terminal device is connected (for example Figure 1A The position B in the clock-in area is the same as the position where the terminal device enters the clock-in area (e.g. Figure 1A There may be a deviation between the position A in the figure, which may cause the time of entering the punch-in area determined by the terminal device and the time when the terminal device actually enters the punch-in area to deviate, resulting in unreasonable timing for the terminal device to push the punch-in service information (for example, Figure 1A Location B is pushed but location A is not pushed), resulting in the problem of incorrect triggering of the check-in service information.
[0078] When pushing service information of other behaviors based on cell positioning technology, there may also be problems such as unreasonable push timing and erroneous push of service information, similar to pushing check-in service information.
[0079] Based on the above analysis, it can be seen that the time error of pushing service information to users is the location of the terminal device accessing the target cell (for example Figure 1A B) to the location where you enter the target area (e.g., the check-in area above) (e.g. Figure 1A If the error caused by this time period can be eliminated, service information can be accurately pushed to the user when the user actually enters the target area.
[0080] Considering that the path that the user takes to the target area is relatively fixed, the location where the target behavior is performed in the target area is relatively fixed, and the user accesses the cell (e.g. Figure 1A 1) to actually entering the target area (e.g. Figure 1AThe distance of this section of the journey (the check-in area) is shorter, and the possibility of being disturbed by other things is smaller in a shorter journey, so the time taken for this section of the journey is relatively fixed. Based on this, the embodiment of the present invention can estimate the time spent by the user in this section of the journey, eliminate the error caused by this time, and thus solve the problem of unreasonable service information push timing and erroneous push.
[0081] In the information push method of the embodiment of the present application, the time from when the terminal device accesses the target cell to when it enters the target area is predicted, and after the time has passed, the service information for the target behavior is pushed. This ensures that when the service information for the target behavior is pushed, the terminal device has already entered the target area, and the user can successfully perform the target behavior according to the pushed service information.
[0082] For example, the target behavior is to check in. Figure 1B As shown, when the terminal device accesses cell 1 at location B, the terminal device does not push the punch-in service information on the interface, but when the terminal device enters the target area, for example, when it actually arrives at location A, the punch-in service information is pushed on the interface. Since the terminal device enters the open area, the user can successfully punch in in the punch-in application at location A according to the pushed punch-in service information.
[0083] In one embodiment, in order to improve the accuracy of time prediction, the time from the terminal device accessing the target cell to entering the target area on different paths can be predicted according to the different paths taken by the user to the target area, so as to push service information for the target behavior after this time.
[0084] Let's take the example of clocking in as the target behavior. Figure 2 , assuming that the user has successfully clocked in at location B and location D using the terminal device before, the terminal device records the cell 4 covering location B and the cell 8 covering location D as the target cells.
[0085] When the user goes to the check-in area via path 1, the terminal device can record the cells accessed in sequence on path 1, for example Figure 2 Among the cells 1, 2, 3, and 4 in the path, when arriving at position A, the cell accessed by the terminal device is switched from cell 3 to cell 4, and cell 4 is determined to be the target cell. The terminal device can predict the time it takes for the user to enter the target area from position A based on the above cells 1, 2, 3, and 4 accessed on path 1, and push the check-in service information after that time.
[0086] When the user goes to the check-in area via path 2, the terminal device can record the cells accessed in sequence on path 2, for example Figure 2Among the cells 5, 6, 7, and 8 in path 2, when arriving at position C, the cell accessed by the terminal device is switched from cell 7 to cell 8, and cell 8 is determined to be the target cell. The terminal device can predict the time it takes for the user to enter the target area from position A based on the above cells 5, 6, 7, and 8 accessed on path 2, and push the check-in service information after that time.
[0087] In order to predict the time from the terminal device accessing the target cell to entering the target area on different paths according to the different paths taken by the user to the target area, in one embodiment, the terminal device can train different time prediction models based on the historical travel data of successfully executing the target behavior, thereby learning the travel patterns of the user when traveling to the target area through the time prediction model, and predicting the time from the terminal device accessing the target cell to entering the target area on a path based on the trained time prediction model.
[0088] In an embodiment of the present application, the terminal device can collect travel data during the trip. If the terminal device performs a target behavior during the trip, the travel data of the successful execution of the target behavior will be recorded as historical travel data for model training; otherwise, the travel data collected during the trip will be deleted.
[0089] In an embodiment of the present application, the historical travel data saved after successfully executing the target behavior may include: information on the cells to which the terminal device has been connected in sequence before the target behavior occurs, cell information on other cells that the terminal device can access before the target behavior occurs, the geographical location of the terminal device when the target behavior occurs, the third cell to which the terminal device is connected when the target behavior occurs, and the first time interval between the time when the target behavior occurs and the first connection to the third cell.
[0090] Optionally, the training and updating of the time prediction model based on historical travel data can be performed periodically. The specific value of the period is not limited in the embodiment of the present application, for example, it can be 1 day, 1 week, etc. When the model training is performed for the first time, the first historical travel data collected within a period of time after the function subscription is used to perform the model training; and when the model is subsequently updated, the travel data that completes the target behavior during the period from the last model training or update to the current time is collected as the second historical travel data, and the time prediction model is updated through the second historical travel data.
[0091] The following takes the clock-in scenario for work as an example to illustrate the implementation of the terminal device in the embodiment of the present application to collect travel data of successful clock-ins.
[0092] like Figure 3 As shown, it is a flow chart of the data collection method provided by an embodiment of the present invention.
[0093] See also Figure 3, the user subscribes to the punch-in service on the terminal device, so that the terminal device can push the punch-in service to the user when arriving at the punch-in area later, prompting the user to punch in.
[0094] Users can subscribe to the punch-in service in the corresponding APP through the application assistant or voice assistant installed on the terminal device. Take the application assistant A as an example. In the interface of A assistant, including various services provided by A assistant for users, users can subscribe to various services through preset operation instructions, such as clicking on the corresponding icon.
[0095] Within a few days after completing the subscription to the punch-in service, the arrival time of the user at the punch-in area may not be predicted, but the punch-in service information may be pushed using the method in the relevant technology, and the user's travel data during this time may be collected as the first historical travel data, and the user's travel pattern may be learned accordingly. Optionally, in order to be more in line with the user's travel pattern when punching in for work, the terminal device may only collect the user's travel data on weekdays. For example, the user's travel data within 10 working days after subscribing to the punch-in service may be used as historical travel data to learn the user's travel pattern.
[0096] Specifically, when the terminal device determines that the user has left home on a weekday, it starts to collect cell information of the cell to which the terminal device is connected. In the process of continuously collecting cell information, the collection of cell information is performed once after each fixed time interval. The specific value of the fixed time interval is not limited in the embodiment of the present application, for example, it may be 5 minutes. The collected cell information includes the cell to which the terminal device is connected, and the cell that the terminal device can access but does not access. The fixed time interval preset when collecting cell information may be, for example, 5 minutes. The fixed time interval may be set by default by the terminal device, or may be set by the user in the user interface provided by Assistant A. The shorter the fixed time interval, the higher the collection accuracy.
[0097] Optionally, the terminal device can establish a user profile based on the user's usage patterns of the terminal device, and determine the community corresponding to the user's home address based on the user profile, thereby determining whether the user leaves home on every weekday, and start collecting community information when the user leaves home.
[0098] In the process of continuously collecting cell information, the terminal device must also determine whether the punch-in is finally completed during the trip to determine whether the collected cell information is valid and needs to be saved. When the punch-in is completed, the geographical location of the punch-in point, the cell information of the third cell currently connected to the terminal device, the time interval between the terminal device connecting to the cell and the current completion of the punch-in, and the cell information collected before the successful punch-in, that is, the cells that have been connected and can be connected but not connected before the successful punch-in, are retrieved and recorded as historical travel data. If the trip does not reach the punch-in location and the punch-in is not completed, the cell information collected during the trip will not be used as data for learning the user's travel patterns, and the cell information collected during the trip will be deleted.
[0099] When judging whether the check-in is completed, the terminal device recognizes the page to judge whether the check-in is completed. When the terminal device recognizes that the check-in success page appears, it is determined that the trip has reached the check-in point and the check-in is completed.
[0100] The following describes the implementation process of training a time prediction model based on collected historical travel data.
[0101] When training a time prediction model based on the collected historical travel data, since the user usually has more than one route to the target area, and the travel patterns of the user to the target area via different routes are also different, if a single time prediction model is used to predict the arrival time of the user on different routes, errors are likely to occur. To solve the above problem, when training the time prediction model, the embodiment of the present invention determines several different routes for the user to go to the target area, and performs differentiated learning for the different routes for the user to go to the target area to obtain several different time prediction models, and selects the model of the corresponding route for prediction during actual prediction to improve the accuracy of the model prediction.
[0102] Corresponding to the above embodiment, Figure 4 FIG. 1 is a flowchart of an information push method provided by an embodiment of the present invention, wherein the specific steps of the method include:
[0103] S401, clustering historical travel data to obtain multiple location clusters.
[0104] Specifically, after completing the function subscription, it is necessary to collect the first historical travel data for a period of time, and complete the training of the time prediction model based on the first historical travel data. After completing the time prediction model training, before the next execution of the time prediction model update, it is necessary to collect the second historical travel data, and execute the time prediction model update based on the second historical travel data.
[0105] The historical travel data includes the geographical location of the terminal device when the target behavior occurs. After the terminal device performs the target behavior, it will perform precise positioning through GPS, Beidou, etc. to determine the location coordinates of the terminal device when the target behavior occurs. In a specific embodiment, when the terminal device completes the check-in in the check-in area, the specific location coordinates of the completed check-in point will be obtained.
[0106] In order to distinguish the routes of users to the target area, clustering can be performed according to the distance between the geographical locations of the terminal devices when the target behavior occurs. When clustering, the longitude and latitude coordinates of each target location are usually used to perform DBSCAN clustering, and the check-in points with a spacing less than the first distance threshold are clustered into a location cluster. Generally, the preset sampling range for clustering is 5 meters, and the minimum sampling number is 3, that is, when the number of target behaviors occurring within a radius of 5 meters is greater than or equal to three times, the at least three location information are clustered into a location cluster.
[0107] Although different target locations can be regarded as different routes chosen by users when heading to the target area, and different time prediction models need to be trained, since the locations where users take the same route and perform target behaviors cannot be exactly the same each time, several relatively similar geographical locations can be regarded as users taking the same route and performing target behaviors at the same target location. There are only slight errors in the distance, which will not cause any actual impact. Therefore, several similar geographical locations can be regarded as the same target location through clustering, and a time prediction model can be trained and used together.
[0108] After clustering is completed, all cells connected to the location information of the target behavior in a location cluster, as well as the cells that can be connected but not connected when the target behavior occurs, are used to determine the cell set corresponding to the location cluster and recorded in the second cell list corresponding to the location cluster. If a cell belongs to two or more location clusters at the same time, this cell is deleted from the location cluster with fewer occurrences and retained in the location cluster with the most occurrences. The second cell sets of all location clusters together form a first cell list, which is pre-stored in the terminal device. The first cell list is used to determine whether the user enters the target cell, and the second cell list is used to determine the corresponding location cluster and the corresponding time prediction model.
[0109] like Figure 5 As shown, it is a schematic diagram of a specific embodiment of a clustering location cluster provided by an embodiment of the present invention, and is still described by taking the clocking-in scenario at work as an example.
[0110] The first historical travel data includes the location information of the terminal device when the target behavior occurs in the target area, that is, the specific location where the terminal device punches in the punch-in area. Figure 5, the terminal device has completed 7 check-ins in the check-in area. Each time the check-in is completed, the terminal device will collect the location information of the check-in point and record it as the check-in point. Clustering is performed based on the location information of the 7 check-in points, and two location clusters are obtained by clustering, location cluster 1 and location cluster 2. Among them, location cluster 1 contains 3 check-in points, and location cluster 2 contains 4 check-in points. The 3 check-in points in location cluster 1 can be regarded as one target location, and the 4 check-in points in location cluster 2 can be regarded as one target location.
[0111] When training the time prediction model in the subsequent steps, two time prediction models need to be trained respectively for location cluster 1 and location cluster 2. The training methods of the two time prediction models are the same.
[0112] like Figure 6 FIG. 1 is a schematic diagram of a location cluster provided by an embodiment of the present invention. Figure 5 The position of cluster 1 is shown.
[0113] Location cluster 1 includes check-in point 1, check-in point 2, and check-in point 3 obtained through clustering. The cells connected by the terminal device when checking in at check-in point 1, check-in point 2, and check-in point 3, as well as the cells that the terminal device can connect to but does not connect to when it is at check-in point 1, check-in point 2, and check-in point 3, are used as the second target cell list of location cluster 1.
[0114] See also Figure 6 , check-in point 1 is located in the coverage of cell 1, and the terminal device accesses cell 1 when checking in, so cell 1 is configured in the cell list of location cluster 1; check-in point 2 is located in the coverage of cell 2, and the terminal device accesses cell 2 when checking in, so cell 2 is configured in the cell list of location cluster 1; check-in point 3 is located in the coverage of both cell 2 and cell 3. Although the terminal device accesses cell 2 when checking in, considering that the terminal device may access cell 3 when it reaches check-in point 3 again in the future, cell 2 and cell 3 are configured in the second cell list of location cluster 1. Therefore, the second cell list corresponding to location cluster 1 includes cell 1, cell 2, and cell 3.
[0115] The target cell lists of each location cluster obtained by clustering are counted to obtain the target cell list corresponding to the entire check-in area. Figure 5 The second cell list corresponding to the location cluster 2 shown is cell 4, cell 5, cell 6 and cell 7, and the first cell list corresponding to the check-in area is cell 1, cell 2, cell 3, cell 4, cell 5, cell 6 and cell 7.
[0116] When the location cluster is updated through the second historical travel data, the location information of the target behavior in the second historical travel data is determined. If the distance between the location information and the geographical location where the target behavior occurs in the existing location cluster is less than the first distance threshold, it is directly divided into the location cluster; if the location information belongs to any existing location cluster, a new location cluster is generated based on it to update the existing location cluster. After the location cluster is updated, the second cell list corresponding to each location cluster and the first cell list are updated accordingly.
[0117] S402: training a time prediction model corresponding to each location cluster using historical travel data.
[0118] Specifically, each historical travel data is divided into a corresponding location cluster according to the geographical location where the target behavior occurred in the historical travel data. According to the historical travel data divided by each location cluster, the time prediction model corresponding to each location cluster is trained. The training method of each time prediction model is the same, that is, to determine the feature data and training labels from the historical travel data, and to perform model training based on the sample data composed of the feature data and the training labels. Among them, in addition to the geographical location of the terminal device when the target behavior occurs, the collected historical travel data also includes the first cell connected when the target behavior occurs, the first time interval between the time when the target behavior occurs and the first connection to the first cell, the cell information to which the terminal device has been connected in sequence within the first time before the target behavior occurs, and the cell information of other cells that the terminal device can access within the first time before the target behavior occurs.
[0119] When determining the characteristic data, the first cell in the historical travel data contained in the current location cluster, the cell information that the terminal device has connected to in sequence within the first time before the target behavior occurs, and the cell information of other cells that the terminal device can access within the first time before the target behavior occurs are encoded to obtain the characteristic data. That is, the route of the terminal device is encoded to obtain the characteristic data.
[0120] Among them, when encoding, it is necessary to determine the cell sequence of the terminal device accessing each cell according to the time sequence of collecting the information of each cell, that is, to determine the specific route. Since the terminal device may be covered by multiple cells in most cases, and may connect to other cells when it is in the same position next time, for example, the location A where the terminal device is located is covered by both cell 1 and cell 2. The terminal device is connected to cell 1 during this trip, and may connect to cell 2 when passing through location A next time. Therefore, in order to determine the full representation method of the route, when determining the cell sequence, the cells accessed by the terminal device and the cells that can be accessed but are not accessed will be arranged and combined, and encoded.
[0121] Among them, since different sections on the route have different importance, for example, when a user just leaves home, the distance to the target area is far away and the remaining time is long. Even if the travel rules of this section are learned, it is impossible to accurately predict the subsequent time. Therefore, in order to reduce the storage pressure and calculation pressure of the terminal device, the cell information collected when the user just leaves home can be appropriately deleted, and only the cell information collected in the most recent first time can be retained. In a specific embodiment, the first time can be set to 30 minutes, that is, it is stipulated that the cell information collected every 5 minutes in the past 30 minutes is retained.
[0122] When determining the training label, the first time interval in the historical travel data contained in the current location is used as the training label, that is, the time interval between the time when the target behavior occurs and the first connection to the first cell is used as the training label.
[0123] The feature data and the corresponding training labels are combined into sample data, and the sample data is divided into training samples and evaluation samples. The model is trained through the training samples, and the model error is estimated and updated through the evaluation samples. The time prediction model can be applied to perform time prediction only after it reaches a certain accuracy.
[0124] When training the model, the time prediction model can regress time points or time intervals. The model type can be a simple statistical model, a machine learning model, or a deep learning model. Simple statistical models are suitable for situations with a small sample size and only allow low computational complexity; machine learning models are suitable for situations with a moderate sample size and allowable moderate computational complexity; deep learning models are suitable for situations with a large sample size and allowable greater computational complexity.
[0125] Among them, the model training performed through the first historical travel data and the model updating method performed through the second historical travel data are the same. When performing model training, it is necessary to ensure that the error of the model is less than a certain threshold before the model training can be considered completed. When performing model update, it is necessary to ensure that the error of the model is less than a preset threshold before the updated model is retained, otherwise the model remains unchanged. The error threshold set for model update should be lower than the error threshold set for model training. Optionally, whether to retain the updated model can also be determined by whether the error is reduced.
[0126] The embodiment of the present invention classifies the locations where the target behavior occurs in the target area, trains the corresponding time prediction models, and predicts the time taken for the last distance on different routes respectively, so as to improve the accuracy of predicting the first time.
[0127] The following still takes the target behavior of clocking in as an example to illustrate the model training.
[0128] like Figure 7 FIG. 1 is a flow chart of a model training method provided by an embodiment of the present invention, in which a terminal device learns a user's travel rules based on historical travel data and trains a time prediction model.
[0129] See also Figure 7 , the terminal device is set to trigger the training or update of the time prediction model every day. Wherein, when the model training is not completed, the time prediction model is trained through the first historical travel data; after the time prediction model training is completed, the time prediction model is updated through the second historical travel data. The terminal device can trigger the training or update of the time prediction model at a preset time, such as 3 o'clock in the evening, or during the idle period when the terminal device is charging and the screen is turned off.
[0130] After the training or update of the trigger time prediction model, clustering is performed based on the geographical location of the check-in points in the collected historical travel data to obtain multiple location clusters. The above clustering method can be, for example, a density-based clustering algorithm (DBSCAN). The historical travel data is assigned to the corresponding location clusters according to the cell information accessed by the geographical location when checking in, and the training of the time prediction model corresponding to each location cluster is performed. Among them, the training method for each time prediction model is the same, and the time prediction models corresponding to different location clusters are used to learn the travel patterns of users arriving at different check-in points.
[0131] When the time prediction model training is specifically performed, feature processing is performed on the historical travel data, that is, feature encoding is performed on the cell information connected when punching in recorded in the historical travel data, and the cell information collected within 30 minutes before punching in, as feature data of the sample data. Label processing is also performed, that is, the time interval between the terminal device accessing the cell and completing the punching in is used as the training label of the sample data. The sample data is divided into training samples and test samples, for example, it can be divided into 80% training samples and 20% test samples, and the time prediction model is trained through the training samples. The error of the time prediction model is evaluated through the test samples. When the error of the time prediction model does not decrease, the time prediction model is kept unchanged; when the error of the time prediction model decreases, the time prediction model is updated until the error of the time prediction model meets the requirements.
[0132] The following describes the method of performing time prediction based on the trained time prediction model and pushing service information.
[0133] like Figure 8 FIG. 1 is a flowchart of an information push method provided by an embodiment of the present invention. The method is used to predict the time to reach a target area and can be applied in various scenarios including clocking in at work. The specific steps of the method include:
[0134] S801, continuously collecting information on the cell to which the terminal device is connected at a preset time interval.
[0135] Specifically, each time the terminal device collects cell information of a cell, the cell information includes the cell information of the cell to which the terminal device accesses when collecting the cell information.
[0136] The movement of the terminal device is determined by whether the cell connected to the terminal device has changed. The terminal device determines the user profile through the user's usage record, and determines the cell corresponding to the user's address based on the user profile. When the cell connected to the terminal device changes from the cell corresponding to the user's address, it is determined that the terminal device has moved and the cell information collection begins.
[0137] S802: When the cell accessed by the terminal device is switched from the first cell to the second cell, and the second cell is the target cell, predict the first time according to the continuously collected cell information.
[0138] Specifically, a first cell list is pre-stored in the terminal device, and the first cell list records cell information of several target cells, which is used to determine whether the cell accessed by the terminal device is the target cell.
[0139] Each time the terminal device collects cell information, it determines whether the cell accessed has changed compared with the last time the cell information was collected. When the cell accessed by the terminal device changes from the first cell to the second cell, it determines whether the second cell is in the preset first cell list. When the second cell is a cell in the first cell list, it determines that the second cell is the target cell. The first cell list is composed of the second cell lists of each location cluster.
[0140] The time when the cell currently connected to the terminal device is determined to be the target cell is used as the prediction timing, and the first time when the terminal device arrives at the target area is predicted based on the continuously collected cell information.
[0141] Among them, when predicting the first time when the terminal device arrives at the target area based on the continuously collected cell information, the first time prediction model corresponding to the currently connected target cell can be determined through the second cell list, and the continuously collected cell information can be input into the first time prediction model to predict the first time when the terminal device arrives at the target area.
[0142] Optionally, when continuously collected cell information is input into the first time prediction model, the collected cell information needs to be feature encoded. Specifically, according to the order in which the terminal device collects the cell information, the cells accessed by the continuously collected terminal devices are encoded into a route feature sequence, and the route feature sequence is input into the first time prediction model to perform time prediction.
[0143] Coding methods include but are not limited to integer coding, one-hot coding, word vector coding, longitude and latitude approximate vector coding, sequence feature library coding, etc.
[0144] In a specific embodiment, the cell connected to the terminal device 20 minutes ago, 15 minutes ago, 10 minutes ago, 5 minutes ago, and currently is cell 1, cell 2, cell 3, cell 4, and cell 5 respectively, and the route feature sequence can be encoded as 0102030405. After each update of the collected cell information, the route feature sequence can be updated. For example, when updating the collected cell information, it is determined that the terminal device is connected to cell 6 after connecting to cell 5, and the route feature sequence is updated to 010203040506.
[0145] S803, pushing service information for the target behavior after the first time.
[0146] Specifically, the terminal device determines that the target area has been reached after the first time, and triggers service information corresponding to the target area for the user to operate. For example, when the target area is a check-in point, after the first time, it is determined that the user has entered the check-in point, and the check-in information is pushed to the user for the user to perform the check-in.
[0147] Within a preset time after the service information is displayed on the display interface, it is detected whether the user will perform the corresponding target behavior according to the pushed service information. When the corresponding target behavior is detected, the data collected in this trip is used as historical travel data for updating the subsequent execution time prediction model.
[0148] The embodiment of the present invention collects the cells connected to the terminal device, determines the route of the terminal device to the target area, and predicts the time required for the user to enter the target cell and finally reach the target area by learning the travel habits and time patterns on the corresponding route, so as to improve the accuracy of the arrival time prediction, ensure that the service information is actually pushed at the appropriate time, and take into account low power consumption and high accuracy at the same time, thereby improving the user experience.
[0149] Optionally, in some embodiments, in order to improve the accuracy of time prediction, when executing S802 to predict the first time based on the continuously collected cell information, other characteristics on the travel route such as the travel mode are also considered to improve the accuracy of the predicted arrival time. Therefore, when executing S801 to continuously collect the cell information connected to the terminal device at a preset time interval, other relevant travel information needs to be collected.
[0150] Specifically, when the terminal device collects the cell information, it will also collect the travel information of this trip. Travel information includes but is not limited to transportation mode information, travel date information, and travel time period information under the current travel date. Among them, travel time information and travel date information can be obtained by the terminal device directly retrieving system information. Transportation mode information includes whether this trip is walking, cycling, driving, taking the bus, taking the subway, etc., which can be identified by the terminal device through sensor data such as acceleration sensors and gyroscope sensors. Optionally, transportation mode information can also be obtained through other APPs when it has data permissions for other APPs. For example, when the terminal device scans the code through the shared bicycle APP to unlock the shared bicycle, it is determined that the transportation mode of this trip is cycling; when the terminal device scans the code through the subway APP to enter the subway station, it is determined that the transportation mode of this trip is taking the subway.
[0151] In a specific embodiment, the user chooses to walk to the target area at 9 am on Monday, and on the way to the target area, the terminal device connects to cell 1, cell 2, cell 3, cell 4, and cell 5 in sequence, and encodes them to obtain a route feature sequence of 0109010102030405.
[0152] Optionally, in some embodiments, before executing S802 to predict the first time based on the continuously collected cell information, it is also necessary to verify the error of the first time prediction model to determine whether the first time prediction model is valid. Only when the first time prediction model is valid, the first time prediction model will be used to predict the first time when the terminal device arrives at the target area.
[0153] Specifically, the model error of the first-time prediction model is determined. When the model error of the first-time prediction model is less than a preset threshold, the first-time prediction model is determined to be valid, and the encoded route feature sequence is input into the first-time prediction model to execute time prediction. When the error of the first-time prediction model is greater than or equal to the preset threshold, the first-time prediction model is determined to be invalid, that is, the first time predicted by the first-time prediction model will have a large error, and the first-time prediction model will no longer be used for prediction. The current moment is determined as the first time for the terminal device to arrive at the target area, and service information is immediately pushed to the user.
[0154] Among them, when evaluating the model error, the evaluation can be performed through the mean absolute error MAE of the first-time prediction model, or the root mean square error RMSE.
[0155] Optionally, when performing encoding of the cell to which the terminal device is connected, feature encoding can be performed by a variety of methods. Specifically, integer encoding assigns a separate integer ID to each cell information. The advantage is that the processing is simple, but the disadvantage is that the cell information is regarded as fixed-distance data, which has a large error; the advantage of one-hot encoding is that the processing is simple, but the disadvantage is that the encoded vector is too sparse, resulting in low efficiency in calculation and storage, and the distance information between the cell information is not used; word vector encoding uses the sequence information of historical cell information, and obtains a low-dimensional word vector for each cell information based on the training mode of the continuous bag-of-words model or the skip-gram model. The advantage is that it can effectively measure the distance between cells and the dimension is low, but the disadvantage is that the processing is more complicated; longitude and latitude approximate vector encoding uses operator public data to map each cell information to the longitude and latitude of the base station, and uses the longitude and latitude as the vector of the cell information. The advantage is that it can effectively measure the distance between cell information and the dimension is very low, but the disadvantage is that the operator public data needs to be used; sequence feature library encoding forms a feature library based on the check-in cell information sequence by clustering, and the similarity measurement adopts dynamic time warping (Dynamic Time Warping). The DTW (Deep Distance Warping) method ensures that the cell information sequences of the same route but different lengths and rhythms of the user can be clustered into the same cluster. Each route feature sequence is assigned to each cluster, and then the clusters are encoded with simple encoding such as one-hot encoding. The advantage is that longer route feature sequences can be compressed into feature library encoding, but the disadvantage is that it is difficult to encode cell information that has not appeared.
[0156] The following is a specific example to illustrate the implementation of the information push method of the embodiment of the present application.
[0157] like Fig. 9 FIG. 1 is a flow chart of an information push method provided by an embodiment of the present invention, which corresponds to a process in which a terminal device performs time prediction and pushes service information.
[0158] See also Fig. 9 , the user of the terminal device leaves home on a weekday and starts to collect the cell information of the cell to which the terminal device is connected. In the process of continuously collecting the cell information, the cell information is collected every 5 minutes. In the process of collecting the cell information, it is determined whether the cell to which the terminal device is connected has changed. When the cell to which the terminal device is connected has changed, it is determined whether the cell newly connected by the terminal device is the target cell.
[0159] When the cell newly accessed by the terminal device is the target cell, determine the second cell list to which the target cell belongs, and determine the time prediction model corresponding to the second cell list as the first time prediction model for actually performing the prediction. Determine whether the error of the first time prediction model is less than the preset threshold. When it is less than the preset threshold, encode the collected cell information, and input the encoding result into the first time prediction model to predict the time when the terminal device arrives at the corresponding punch-in point. After the predicted time has passed, assistant A pushes the punch-in service to the user on the main interface of the terminal device for the user to perform the punch-in. When the error of the first time prediction model is greater than or equal to the preset threshold, it is determined that the model is unavailable, and the punch-in service is immediately pushed to the user through assistant A for the user to perform the punch-in.
[0160] Among them, the check-in service includes the icon of a third-party check-in application as a link. Users can quickly enter the check-in interface of the check-in application by clicking the icon of the check-in application, and complete preset operation instructions such as check-in in the check-in application.
[0161] like Fig.10 The figure is a schematic diagram of the structure of a terminal device provided by the present invention.
[0162] See also Fig.10 The terminal device includes an application layer and an application framework layer. The application layer includes an A assistant, and the application framework layer includes a processing module, a perception module, a data middle platform, and a learning middle platform.
[0163] A assistant is generally an APP installed in the terminal device, which is used to push service information.
[0164] The perception module specifically includes a geo-fence perception module, a data collection perception module, and a page perception module, which are used to perceive and obtain various types of data and send the obtained data to the data center for processing. Specifically, the geo-fence perception module is used to determine the cell to which the terminal device is connected, the data collection module is used to continuously collect cell information, and the page perception module is used to identify the interface of the terminal device.
[0165] The processing module is used to instruct the learning center to train a time prediction model based on the data in the data center, and predict the first time based on the time prediction model.
[0166] like Fig.11 As shown, it is a flowchart of a specific embodiment of the information push method provided by the present invention. Fig.11 The interaction sequence diagram of each functional component is shown.
[0167] See also Fig.11After the user subscribes to the punch-in service on the corresponding page, the page perception module will identify the interface of successful subscription and send the result of successful subscription to the processing module. The processing module will push the punch-in information to the user when he arrives at the punch-in point later, and save the collected community information when the punch-in is successful as historical travel data. And subscribe to the away fence to sense whether the community connected to the terminal device has changed. After subscribing to the away fence, the geographic fence module senses the change between the accessed community and the community that the address should access, so as to determine that the user holding the terminal device has left home, and collects the accessed community information every 5 minutes through the data collection perception module, and saves the collected data to the data center.
[0168] The user subscribes to the clock-in fence, and the processing module triggers the execution time prediction when arriving at the target cell connected to the target area. The geo-fence module performs positioning to determine the arrival at the clock-in point, that is, the arrival at the target cell, determines the corresponding location cluster based on the currently connected target cell, and determines the first-time prediction model of the location cluster from the learning center. The learning center determines the model error of the first-time prediction model to determine whether the first-time prediction model is valid. When the model fails, the push module immediately pushes the clock-in reminder to the user. When the model takes effect, the first-time prediction model is used to predict the clock-in time based on the collected cell information, and the push module pushes the clock-in reminder to the user after the predicted clock-in time.
[0169] After the user successfully clocks in through the corresponding interface, the page perception module identifies the corresponding page, and collects the cell information to which the terminal device is connected when the clock-in is successful, the longitude and latitude coordinates of the terminal device when the clock-in is successful, and calculates the time interval from the terminal device connecting to the target cell to completing the clock-in. The collected data is stored in the data center, and the data is divided into two parts, namely training samples and evaluation samples.
[0170] The learning platform trains and updates the time prediction model. The check-in points are clustered by longitude and latitude coordinates to obtain several location clusters. The collected historical travel data is divided into corresponding location clusters to train and update each time prediction model.
[0171] The above embodiment can be extended to other scenarios. The check-in area can be extended to any target area that the user wants to reach. The check-in behavior performed in the check-in area can be any target behavior related to the target area performed by the user in the target area. The pushed check-in service can be the service information corresponding to the target behavior.
[0172] In a specific embodiment, when a user holds a terminal device and goes to a subway station to take a train, the above information push method can be applied. The target area can be a subway station where the user often takes a train, and the target behavior corresponding to the target area is to show the subway boarding code. When the terminal device arrives at the subway station gate, it pushes the subway APP to help the user quickly show the subway boarding code.
[0173] Specifically, after the user subscribes to the boarding code service through the A assistant interface, the terminal device can determine the subway station that the user often goes to by taking the subway through the geographic location collected when taking the subway in the past, and use the subway station or the gate in the station as the target location. When the user leaves home, collect the cell information connected to the user during the travel route, as well as the cell information that can be accessed but not connected, and when the user opens the subway APP through the terminal device and retrieves the subway boarding code, it is determined that the user has performed the target behavior, and collect the geographic location of the terminal device at this time, the first cell connected at this time, and the first time interval between the first connection to the first cell and the retrieval of the subway boarding code. The collected data is used as history for training the time prediction model.
[0174] After the model training is completed, when the user leaves home, the user's travel route is determined by collecting the cell information connected to the terminal device, and the time prediction model is used to predict the first time the user enters the target cell and arrives at the subway gate. When the first time is reached, the subway service is displayed on the display interface of the terminal device, and the user can click on the subway service link to enter the subway APP's boarding code interface to quickly show the boarding code.
[0175] In another specific embodiment, when a user holds a terminal device and goes to a courier station to pick up a courier, the above information push method can be applied. The target area can be a courier station. The target behavior corresponding to the target area is to present the courier pickup code. When the terminal device arrives at the courier station, it pushes a text message APP to help the user quickly find the courier pickup code.
[0176] Specifically, after the user subscribes to express pickup through the A assistant interface, the terminal device can determine the express station that the user often goes to through the geographic location collected by the user in the past interval, and use the express station as the target location. When the user leaves home, collect the cell information connected to the user during the travel route, as well as the cell information that can be accessed but not connected, and when the user opens the SMS APP through the terminal device and retrieves the express pickup code, determine that the user has performed the target behavior, collect the geographic location of the terminal device at this time, the first cell connected at this time, and the first time interval between the first connection to the first cell and the retrieval of the express pickup code. The collected data is used as history for training the time prediction model.
[0177] After the model training is completed, when the user leaves home, the user's travel route is determined by collecting the cell information connected to the terminal device, and the time prediction model is used to predict the first time from the user entering the target cell to the courier station. When the first time is reached, the courier pickup service is displayed on the display interface of the terminal device. The user can click on the courier pickup service link to enter the pickup code interface of the SMS APP to quickly query the pickup code.
[0178] like Fig.12 FIG. 1 is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present invention. The electronic device can be implemented as a terminal device provided in an embodiment of the present application. Fig.12 As shown, the electronic device 1200 may include a processor 1201, an internal memory 1202, an antenna 1, an antenna 2, a mobile communication module 1203, a wireless communication module 1204, a display screen 1205, and the like.
[0179] It is to be understood that the structure illustrated in the embodiment of the present application does not constitute a specific limitation on the electronic device 1200. In other embodiments of the present application, the electronic device 1200 may include more or fewer components than shown in the figure, or combine certain components, or split certain components, or arrange the components differently. For example, in a computer device, there may be no antenna, mobile communication module, and wireless communication module. The illustrated components may be implemented in hardware, software, or a combination of software and hardware.
[0180] The processor 1201 of the electronic device 1200 may be a device on a chip (SOC), which may include a central processing unit (CPU) and may further include other types of processors. For example, the processor 1201 may include an application processor (AP) and / or a neural-network processing unit (NPU).
[0181] The processor 1201 may include one or more processing units. Different processing units may be independent components or integrated into one or more processors. In some embodiments, the electronic device 1200 may also include one or more processors 1201. The controller may generate an operation control signal according to the instruction operation code and the timing signal to complete the control of fetching and executing instructions.
[0182] NPU is a neural network (NN) computing processor. By drawing on the structure of biological neural networks, such as the transmission mode between neurons in the human brain, it can quickly process input information and continuously self-learn. NPU can realize applications such as intelligent cognition of electronic devices, such as image recognition, face recognition, voice recognition, text understanding, etc.
[0183] The internal memory 1202 of the electronic device 1200 can be used to store one or more computer programs, which include instructions. The processor 1201 can run the above instructions stored in the internal memory 1202, so that the electronic device 1200 performs the methods provided in some embodiments of the present application, as well as various applications and data processing. The internal memory 1202 may include a code storage area and a data storage area. Among them, the code storage area can store an operating system. The data storage area can store data created during the use of the electronic device 1200, etc. In addition, the internal memory 1202 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more disk storage components, flash memory components, universal flash storage (UFS), etc.
[0184] The internal memory 1202 can be a read-only memory (ROM), other types of static storage devices that can store static information and instructions, a random access memory (RAM) or other types of dynamic storage devices that can store information and instructions, or an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical disc, laser disc, optical disc, digital versatile disc, Blu-ray disc, etc.), a magnetic disk storage medium or other magnetic storage device, or it can also be any computer-readable medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer.
[0185] The processor 1201 and the internal memory 1202 can be combined into a processing device, or more commonly, are independent components. The processor 1201 is used to execute the program code stored in the internal memory 1202 to implement the method described in the embodiment of the present application. In specific implementation, the internal memory 1202 can also be integrated into the processor, or independent of the processor.
[0186] The wireless communication function of the electronic device 1200 can be implemented through antenna 1, antenna 2, mobile communication module 1203, wireless communication module 1204, modem processor and baseband processor.
[0187] Antenna 1 and antenna 2 are used to transmit and receive electromagnetic wave signals. Each antenna in the electronic device 1200 can be used to cover a single or multiple communication frequency bands. Different antennas can also be reused to improve the utilization of the antennas. For example, antenna 1 can be reused as a diversity antenna for a wireless local area network. In some other embodiments, the antenna can be used in combination with a tuning switch.
[0188] The mobile communication module 1203 can provide solutions for wireless communications including 2G / 3G / 4G / 5G, etc., applied to the electronic device 1200. The mobile communication module 1203 may include at least one filter, a switch, a power amplifier, a low noise amplifier (LNA), etc. The mobile communication module 1203 can receive electromagnetic waves from the antenna 1, and filter, amplify, and process the received electromagnetic waves, and transmit them to the modulation and demodulation processor for demodulation. The mobile communication module 1203 can also amplify the signal modulated by the modulation and demodulation processor, and convert it into electromagnetic waves for radiation through the antenna 1. In some embodiments, at least some functional modules of the mobile communication module 1203 can be set in the processor 1201.
[0189] The modem processor may include a modulator and a demodulator. The modulator is used to modulate the low-frequency baseband signal to be sent into a medium-high frequency signal. The demodulator is used to demodulate the received electromagnetic wave signal into a low-frequency baseband signal. The demodulator then transmits the demodulated low-frequency baseband signal to the baseband processor for processing. After the low-frequency baseband signal is processed by the baseband processor, it is passed to the application processor. The application processor displays an image or video through the display screen 1205. In some embodiments, the modem processor may be an independent device. In other embodiments, the modem processor may be independent of the processor 1201 and be set in the same device as the mobile communication module 1203 or other functional modules.
[0190] The wireless communication module 1204 can provide wireless communication solutions including wireless local area networks (WLAN) (such as wireless fidelity (Wi-Fi) networks), bluetooth (BT), global navigation satellite system (GNSS), frequency modulation (FM), near field communication (NFC), infrared (IR), etc., applied to the electronic device 1200. The wireless communication module 1204 can be one or more devices integrating at least one communication processing module. The wireless communication module 1204 receives electromagnetic waves via antenna 2, modulates and filters the electromagnetic wave signals, and sends the processed signals to the processor 1201. The wireless communication module 1204 can also receive the signal to be sent from the processor 1201, modulate the frequency, amplify it, and convert it into electromagnetic waves for radiation through antenna 2. In the embodiment of the present application, the communication connection between the anchor device and the sub-device can be a Wi-Fi network provided by the wireless communication module 1204.
[0191] In some embodiments, antenna 1 of electronic device 1200 is coupled to mobile communication module 1203, and antenna 2 is coupled to wireless communication module 1204, so that electronic device 1200 can communicate with the network and other devices through wireless communication technology.
[0192] The electronic device 1200 can implement touch input through the display screen 1205, the touch sensor, and the processor 110. For example, the touch sensor and the display screen 1205 are integrated into a touch screen, and the user's click operation on the touch screen is collected by the touch sensor as a touch signal, and the touch signal is collected and converted by the sensor module 180 and transmitted to the processor 110, and the processor 110 recognizes the touch signal and analyzes it into the user's touch operation behavior.
[0193] Furthermore, the devices, apparatuses, and modules described in the embodiments of the present application may be implemented by computer chips or entities, or by products having certain functions.
[0194] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, devices, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code.
[0195] In the several embodiments provided in the present application, if any function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application.
[0196] Specifically, in one embodiment of the present application, a computer-readable storage medium is further provided, in which a computer program is stored. When the computer program is run on a computer, the computer executes the method provided in the embodiment of the present application.
[0197] An embodiment of the present application further provides a computer program product, which includes a computer program. When the computer program is run on a computer, the computer executes the method provided by the embodiment of the present application.
[0198] The embodiment description in the present application is described with reference to the flowchart and / or block diagram of the method, equipment (device) and computer program product according to the embodiment of the present application. It should be understood that each flow process and / or box in the flow chart and / or block diagram and the combination of the flow chart and / or box in the flow chart and / or block diagram can be realized by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processing machine or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device produce a device for realizing the function specified in one flow chart or multiple flows and / or one box or multiple boxes of the block diagram.
[0199] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction device that implements the functions specified in one or more processes in the flowchart and / or one or more boxes in the block diagram.
[0200] These computer program instructions may also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, whereby the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one or more processes in the flowchart and / or one or more boxes in the block diagram.
[0201] It should also be noted that in the embodiments of the present application, "at least one" refers to one or more, and "more than one" 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 the existence of A alone, the existence of A and B at the same time, and the existence of B alone. Among them, A and B can be singular or plural. The character " / " generally indicates that the previous and subsequent associated objects are in an "or" relationship. "At least one of the following" and similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b and c can be represented by: a, b, c, a and b, a and c, b and c, or a and b and c, where a, b, c can be single or multiple.
[0202] In the embodiments of the present application, the terms "include", "comprises" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the presence of other identical elements in the process, method, commodity or device including the element.
[0203] The present application may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules may be located in local and remote computer storage media, including storage devices.
[0204] Each embodiment in this application is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.
[0205] Those skilled in the art will appreciate that the various units and algorithm steps described in the embodiments of the present application can be implemented by a combination of electronic hardware, computer software, and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.
[0206] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described devices, apparatuses and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.
[0207] The above is only a specific implementation of the present application. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. The protection scope of the present application should be based on the protection scope of the claims.
Claims
1. An information push method, characterized in that: Applied to a terminal device, the method comprises: Continuously collect cell information accessed by the terminal device at preset time intervals; When the cell accessed by the terminal device switches from the first cell to the second cell, and the second cell is the target cell, the continuously collected cell information is input into the first time prediction model, and the duration output by the first time prediction model is used as the first time; the first time prediction model is used to predict the duration according to the continuously collected cell information; the first time prediction model is a time prediction model corresponding to a second cell list to which the second cell belongs, the second cell list includes cells covering geographical locations in a first location cluster, the first location cluster is a location cluster corresponding to the second cell list to which the second cell belongs, and each of the location clusters includes geographical locations where the user is located when the target behavior is successfully performed, and the distance between each other is less than a first distance threshold; After the first time, service information for the target behavior is displayed.
2. The method according to claim 1, characterized in that The second cell is a target cell, including: The second cell is a cell in the first cell list; The first cell list is used to record cells covering the geographical location where the user is located when the target behavior successfully occurs, and the target behavior is a user operation instruction corresponding to the service information.
3. The method according to claim 1, characterized in that The method further comprises: Collect at least one of the following information: Transportation information, travel date information, and travel time period information under the current travel date; The collected information is used to be input into the first time prediction model together with the cell information.
4. The method according to claim 2, characterized in that: The second cell list corresponding to each of the location clusters constitutes the first cell list.
5. The method according to any one of claims 1 to 4, characterized in that: Before inputting the continuously collected cell information into the first time prediction model, the method further includes: Determine whether a model error of the first time prediction model is less than a preset threshold.
6. The method according to claim 5, characterized in that Also includes: When it is determined that the model error of the first time prediction model is not less than a preset threshold, the service information is displayed.
7. The method according to any one of claims 1 to 4, characterized in that: The step of inputting the continuously collected cell information into a first time prediction model comprises: According to the order of collection time, the continuously collected cell information is encoded into a route feature sequence; The route feature sequence is input into the first time prediction model.
8. The method according to any one of claims 1 to 4, characterized in that: After displaying the service information for the target behavior after the first time, the method further includes: When the target behavior is detected, the cell information continuously collected by the terminal device, the geographical location of the terminal device when the target behavior occurs, the third cell to which the terminal device is connected when the target behavior occurs, the first time interval between the time when the target behavior occurs and the time when the first connection to the third cell occurs, and the cell information of other cells accessible to the terminal device before the target behavior occurs are recorded as the historical travel data of the target behavior.
9. The method according to any one of claims 1 to 4, characterized in that: The method further comprises: Acquire first historical travel data; the first historical travel data is used to record historical travel data of the target behavior that has occurred; Clustering the first historical travel data to obtain a location cluster; A second cell list corresponding to each location cluster is determined according to the first historical travel data corresponding to each location cluster, and a time prediction model corresponding to each location cluster is trained.
10. The method according to claim 9, characterized in that The first historical travel data includes: The geographical location of the terminal device when the target behavior occurs, the third cell connected when the target behavior occurs, other cells that can be accessed when the target behavior occurs, the first time interval between the time when the target behavior occurs and the time when it is first connected to the third cell, the cell information to which the terminal device has been connected in sequence before the target behavior occurs, and the cell information of other cells that the terminal device can access before the target behavior occurs.
11. The method according to claim 9, characterized in that The step of clustering the first historical travel data to obtain a location cluster includes: According to the distances between the geographical locations where the target behavior occurred in the first historical travel data, the geographical locations whose distances are less than the first distance threshold are clustered to obtain a plurality of location clusters.
12. The method according to claim 9, characterized in that The determining, according to the first historical travel data corresponding to each of the location clusters, a second cell list corresponding to each of the location clusters comprises: The second cell list is determined according to the third cell connected when the target behavior occurs in the first historical travel data corresponding to each of the location clusters, and other cells accessible when the target behavior occurs.
13. The method according to claim 9, characterized in that Training a time prediction model corresponding to each of the location clusters according to the first historical travel data corresponding to each of the location clusters includes: For the first historical travel data of each target behavior in each location cluster, encode the third cell in the first historical travel data, the cell information that the terminal device has sequentially connected to before the target behavior occurs, and the cell information of other cells that the terminal device can access before the target behavior occurs to obtain feature data; using the first time interval in the first historical travel data as a training label; Combining the feature data and the corresponding training labels into sample data; Dividing the sample data of each of the location clusters into a training sample and an evaluation sample for each of the location clusters; A time prediction model corresponding to each of the location clusters is obtained by training according to the training samples and evaluation samples of each of the location clusters.
14. The method according to claim 10, characterized in that Also includes: Acquire second historical travel data, where the second historical travel data is used to record historical travel data of the target behavior that has occurred after the previous time prediction model training; The second cell list is updated according to the second historical travel data, and the time prediction model is retrained.
15. The method according to claim 14, characterized in that Retraining the time prediction model according to the second historical travel data includes: For the geographical locations where the target behavior occurred in the second historical travel data, geographical locations whose distances to geographical locations in the existing location clusters are less than a first distance threshold are assigned to the existing location clusters, and geographical locations that are not assigned to the existing location clusters are clustered to obtain new location clusters; Update and train the time prediction model corresponding to the existing location cluster according to the second historical travel data corresponding to the existing location cluster; The time prediction model corresponding to the new location cluster is trained according to the second historical travel data corresponding to the new location cluster.
16. The method according to claim 15, characterized in that The method further comprises: If the model error of the time prediction model obtained by retraining is less than a preset threshold, updating the corresponding time prediction model; If the model error of the time prediction model obtained by retraining is greater than or equal to a preset threshold, the corresponding time prediction model is kept unchanged.
17. An electronic device, characterized in that: The electronic device comprises a memory for storing computer program instructions and a processor for executing the computer program instructions, wherein when the computer program instructions are executed by the processor, the electronic device is triggered to execute the method according to any one of claims 1 to 16.
18. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which, when executed on a computer, enables the computer to execute the method according to any one of claims 1 to 16.
