Interest position point determination method and device, terminal equipment and storage medium

Abstract

The embodiment of the invention discloses an interest position point determination method and device, terminal equipment and a storage medium, the method is applied to the terminal equipment, and the method comprises the following steps: obtaining log data generated by the terminal equipment between a previous interest position point and a current position point; generating a predicted log event sequence according to the log data; inputting the predicted log event sequence into a trained interest position point inference model, and determining a next interest position point of the terminal equipment according to the predicted log event sequence through the interest position point inference model; the interest position point inference model is obtained by training according to a plurality of sample log event sequences and termination interest position points corresponding to the sample log event sequences. By implementing the embodiment of the invention, the timeliness and accuracy of determining the interest position point which the terminal device is about to arrive at can be improved.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a method, apparatus, terminal device and storage medium for determining a point of interest. Background Art

[0002] Points of Interest (POI) can be interesting or meaningful places on a map, such as stores, hospitals, schools, etc. The terminal device senses whether the user is about to arrive at a frequently visited point of interest (POI), such as home or company, which is one of the key points to improve the user experience. At present, the method for determining points of interest is usually to use the Global Positioning System (GPS) to locate, calculate the geographical distance between the current position and the POI, and determine that the user has arrived at the POI if the geographical distance is less than a certain set threshold. However, continuous recording of location information through GPS will cause the power consumption of the terminal device to be too high. Therefore, passive location recording is usually performed every five minutes, but this will cause a lag in the perception of points of interest, and it is impossible to determine the points of interest that the user is about to arrive in time, so that it is impossible to recommend appropriate services to the user in time. For example, the user is reminded to pick up the package only after the user has arrived home, and the gourmet restaurant is recommended only after the user has started dining. Therefore, how to determine the points of interest that are about to arrive in a timely and accurate manner has become a technical problem that needs to be solved urgently. Summary of the invention

[0003] The embodiments of the present application disclose a method, an apparatus, a terminal device, and a storage medium for determining a point of interest, which can improve the timeliness and accuracy of determining a point of interest that a terminal device is about to reach.

[0004] The embodiments of the present application also disclose a training method, apparatus, terminal device and storage medium for an interest location point reasoning model, which can improve the accuracy of the interest location point reasoning model, thereby improving the accuracy of using the interest location point reasoning model to predict the interest location point that the terminal device is about to reach.

[0005] The present application embodiment discloses a method for determining a point of interest, which is applied to a terminal device, and the method includes:

[0006] Obtaining log data generated by the terminal device between the last point of interest and the current point of interest;

[0007] generating a predicted log event sequence according to the log data;

[0008] The predicted log event sequence is input into the trained interest location point inference model to determine the next interest location point of the terminal device according to the predicted log event sequence through the interest location point inference model; the interest location point inference model is trained based on multiple sample log event sequences and the terminal interest location points corresponding to each of the sample log event sequences.

[0009] The present application embodiment discloses a method for training an interest point inference model, the method comprising:

[0010] Obtain historical log data of terminal devices;

[0011] Generating a plurality of sample log event sequences between any two preset interest locations according to the historical log data;

[0012] The interest point inference model to be trained is trained according to a plurality of sample log event sequences and the termination interest points corresponding to each of the sample log event sequences.

[0013] The present application embodiment discloses a device for determining a point of interest, which is applied to a terminal device, and the device includes:

[0014] A first acquisition module, used to acquire log data generated by the terminal device between the last interested location point and the current location point;

[0015] A first generating module, used to generate a predicted log event sequence according to the log data;

[0016] An inference module is used to input the predicted log event sequence into a trained interest point inference model, so as to determine the next interest point of the terminal device according to the predicted log event sequence through the interest point inference model; the interest point inference model is trained based on multiple sample log event sequences and the terminal interest point corresponding to each of the sample log event sequences.

[0017] The present application embodiment discloses a training device for an interest point inference model, the device comprising:

[0018] The second acquisition module is used to acquire historical log data of the terminal device;

[0019] A second generating module is used to generate a plurality of sample log event sequences between any two preset interest locations according to the historical log data;

[0020] The first training module is used to train the interest point inference model to be trained according to a plurality of sample log event sequences and the terminal interest point corresponding to each sample log event sequence.

[0021] An embodiment of the present application discloses a terminal device, including a memory and a processor, wherein a computer program is stored in the memory, and when the computer program is executed by the processor, the processor implements a method in any embodiment disclosed in the embodiment of the present application.

[0022] The embodiments of the present application disclose a computer-readable storage medium storing a computer program, wherein the computer program enables a computer to execute a method in any embodiment disclosed in the embodiments of the present application.

[0023] Compared with the related art, the embodiment of the present application discloses a method, apparatus, terminal device and storage medium for determining a point of interest, which has the following beneficial effects:

[0024] A predicted log event sequence is generated based on the log data generated by the terminal device between the previous interest point and the current position point; the interest point inference model obtained through training determines the next interest point of the terminal device based on the input predicted log event sequence; the interest point inference model is trained based on multiple sample log event sequences and the terminal interest point corresponding to each sample log event sequence.

[0025] The embodiment of the present application obtains the log data of the terminal device and predicts the next location point of interest of the terminal device, thereby reducing the power consumption of the terminal device caused by real-time positioning through GPS; and, based on the log data generated by the terminal device between the previous location point of interest and the current location point, it is possible to generate a predicted log event sequence in advance before the terminal device reaches the next location point of interest. The generated predicted log event sequence can accurately reflect the user's true behavior and intention, and determine the next location point of interest of the terminal device according to the input predicted log event sequence through the interest location point reasoning model, so that the next location point of interest to be reached by the user can be determined in advance when the user has not yet reached the location point of interest, which can improve the timeliness and accuracy of determining the location point of interest that the terminal device is about to reach.

[0026] The present application also discloses a method, apparatus, terminal device and storage medium for training an interest point inference model, which has the following beneficial effects:

[0027] According to the historical log data of the terminal device, a plurality of sample log event sequences between any two preset interest locations are generated; and based on the plurality of sample log event sequences, and the termination interest locations corresponding to each sample log event sequence, the interest location point reasoning model to be trained is trained. The embodiment of the present application trains the interest location point reasoning model to be trained through the plurality of sample log event sequences generated by the historical log data and the termination interest locations corresponding to each sample log event sequence, and does not rely solely on the location data of the terminal device, so that the sample data used to train the interest location point reasoning model is more comprehensive and accurately reflects the user's real behavior and intention, and can improve the accuracy of the interest location point reasoning model, thereby improving the accuracy of using the interest location point reasoning model to predict the interest location point that the terminal device is about to reach. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the technical solutions in 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.

[0029] Figure 1 It is a schematic diagram of an application scenario of a method for determining a location point of interest disclosed in an embodiment of the present application;

[0030] Figure 2 It is a flowchart of a method for determining a point of interest location disclosed in an embodiment of the present application;

[0031] Figure 3 is a schematic diagram of a clustering result disclosed in an embodiment of the present application;

[0032] Figure 4 is a flowchart of another method for determining a point of interest location disclosed in an embodiment of the present application;

[0033] Figure 5a is a schematic diagram of a complete log event sequence disclosed in an embodiment of the present application;

[0034] Figure 5b is a schematic diagram of a sample log event sequence intercepted from a complete log event sequence disclosed in an embodiment of the present application;

[0035] Figure 6 It is a flowchart of a method for training an interest location point inference model disclosed in an embodiment of the present application;

[0036] Figure 7 It is a structural diagram of an interest location point reasoning model disclosed in an embodiment of the present application;

[0037] Figure 8 It is a schematic diagram of the reasoning process of an interest location point reasoning model disclosed in an embodiment of the present application;

[0038] Fig. 9 It is a structural schematic diagram of a device for determining a point of interest location disclosed in an embodiment of the present application;

[0039] Fig.10 It is a structural schematic diagram of a training device for an interest location point reasoning model disclosed in an embodiment of the present application;

[0040] Fig.11 It is a structural diagram of a terminal device disclosed in an embodiment of the present application. DETAILED DESCRIPTION

[0041] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, 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 this field without creative work are within the scope of protection of this application.

[0042] It should be noted that the terms "including" and "having" and any variations thereof in the embodiments of the present application and the accompanying drawings are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device including a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products or devices.

[0043] The embodiments of the present application disclose a method, an apparatus, a terminal device, and a storage medium for determining a point of interest, which can improve the timeliness and accuracy of determining a point of interest that a terminal device is about to reach.

[0044] The embodiments of the present application also disclose a training method, apparatus, terminal device and storage medium for an interest location point reasoning model, which can improve the accuracy of the interest location point reasoning model, thereby improving the accuracy of using the interest location point reasoning model to predict the interest location point that the terminal device is about to reach.

[0045] The following are detailed descriptions of each.

[0046] See also Figure 1 , Figure 1 Schematic diagram of an application scenario of a method for determining a point of interest location disclosed in an embodiment of the present application. Figure 1As shown, the user and the terminal device 10 can move together. The terminal device 10 may include but is not limited to mobile phones, wearable devices, tablet computers, vehicle-mounted terminals and other terminal devices, which are not limited in the embodiments of the present application.

[0047] The last point of interest of the terminal device 10 is point A, the next point of interest to be reached by the terminal device is point B, and the current point of the terminal device 10 is point C. The terminal device can obtain the log data generated by the terminal device between point A and point C, and generate a predicted log event sequence based on the log data; through the point of interest inference model, according to the input predicted log event sequence, it is determined in advance that the user is going to reach point B.

[0048] It should be noted that the point of interest in the embodiment of the present application is the point of interest POI. Optionally, the point of interest can be represented by latitude and longitude values.

[0049] Please read further Figure 2 , Figure 2 is a flow chart of a method for determining a point of interest location disclosed in an embodiment of the present application, which method can be applied to a terminal device; Figure 2 As shown, the method for determining the point of interest location may include the following steps:

[0050] 201. Obtain log data generated by the terminal device between the last point of interest and the current point of interest.

[0051] The last point of interest is the last point of interest reached by the terminal device; the current point of interest is the current point of interest of the terminal device.

[0052] Log data is a record generated on a terminal device, including information about various activities and events of the operating system, application programs, and the device. In some embodiments, the log data includes at least one of wireless connection data, behavior recognition data, location data, and application usage data.

[0053] The wireless connection data may include but is not limited to Bluetooth log data, Wireless Fidelity (WiFi) log data, etc. Bluetooth log data can be used to record the interaction between the terminal device and other Bluetooth devices, including Bluetooth connection, Bluetooth disconnection, etc.; WiFi log data can be used to record the interaction between the terminal device and the wireless network, including WiFi connection, WiFi disconnection, etc.

[0054] Behavior recognition data can be used to record data collected by the device sensors of the terminal device. The device sensors may include but are not limited to accelerometers, gyroscopes, barometers, etc. The data collected by the device sensors can be used to determine the motion state of the terminal device, such as walking state, running state, elevator state, etc.

[0055] The location data may include but is not limited to base station location data, GPS location data, wireless network location data, etc.

[0056] Base station positioning data is the approximate location data of the terminal device determined based on the signal transmission between the base station and the terminal device. Base station positioning has low power consumption, but the accuracy is also low. Optionally, the terminal device can perform base station positioning according to a preset sampling interval, for example, recording base station positioning data every 5 minutes.

[0057] GPS positioning data is the precise location data of the terminal device determined by a radio navigation positioning system based on artificial earth satellites. GPS positioning consumes high power but has high accuracy. Optionally, the terminal device can perform GPS positioning when the user actively requests the GPS positioning service, for example, when the user turns on the positioning service of the terminal device or when the user opens a navigation application.

[0058] Wireless network positioning data is the location data of terminal devices determined by analyzing wireless network signals, such as WiFi signals, Bluetooth signals, etc.

[0059] Application usage data can be used to record application activity data, such as application startup, login, shutdown, interaction, and other data.

[0060] 202. Generate a predicted log event sequence based on the log data.

[0061] The predicted log event sequence is a log event sequence used to be input into the trained interest point inference model to predict the next interest point. The predicted log event sequence is generated based on the log data generated by the terminal device between the previous interest point and the current location point.

[0062] A log event sequence is a series of log events recorded in chronological order. The terminal device can generate corresponding log events based on the log data.

[0063] For example, assuming that the last location point of interest of the terminal device is the user's home, at 07:30 a.m., the terminal device performs base station positioning at home, and the terminal device records the location data and generates a corresponding log event as a base station positioning event; thereafter, at 07:35 a.m., the terminal device disconnects the home WiFi, and the terminal device records the wireless connection data and generates a corresponding log event as a WiFi disconnection event; at 07:40 a.m., the user takes the elevator with the terminal device, and the terminal device records the behavior recognition data, and determines that the corresponding log event is an elevator event based on the acceleration change data and the ambient air pressure change data in the behavior recognition data; at 07:45 a.m., the user walks with the terminal device, and the terminal device records the behavior recognition data and generates a corresponding log event as a walking event. Event; at 07:55 a.m., the user starts the subway application to open the QR code for boarding, and aligns the QR code for boarding with the scanning port of the subway gate to enter the gate. The terminal device records the application usage data and generates a corresponding log event as a subway event; at 08:00 a.m., the terminal device starts the music application to listen to music, and the terminal device records the application usage data and generates a corresponding log event as a music application usage event; at 08:10 a.m., the terminal device starts the social application, and the terminal device records the application usage data and generates a corresponding log event as a social application usage event; at 08:15 a.m., the terminal device performs GPS positioning, obtains the current location of the terminal device, records the location data and generates a corresponding log event as a GPS positioning event.

[0064] It can be seen that in the above example, the terminal device generates a predicted log event sequence based on the log data generated between the previous location of interest and the current location: base station positioning event, WiFi disconnection event, elevator event, walking event, subway event, music event, GPS positioning event.

[0065] 203. Input the predicted log event sequence into the trained interest point inference model, so as to determine the next interest point of the terminal device according to the predicted log event sequence through the interest point inference model.

[0066] The interest point inference model is trained based on multiple sample log event sequences and the terminal interest point corresponding to each sample log event sequence. The sample log event sequence is a log event sequence used to train the interest point inference model, and the terminal interest point corresponding to the sample log event sequence refers to the interest point reached by the sample log event sequence.

[0067] The next point of interest is the next point of interest to be reached by the terminal device.

[0068] In some embodiments, the sample log event sequence can be generated based on the historical log data of the terminal device. In the process of training the interest point inference model, the historical log data of the terminal device can be obtained, and a plurality of sample log event sequences between any two preset interest points can be generated based on the historical log data, and then the interest point inference model to be trained is trained based on the plurality of sample log event sequences and the terminal interest points corresponding to each sample log event sequence.

[0069] The historical log data may be log data generated by the terminal device in the past period of time, for example, log data generated by the terminal device in the past three months.

[0070] The preset interest points may include points input by the user, or may include points frequently visited by the user determined by the terminal device based on historical log data, without specific limitation.

[0071] The historical log data may include multiple preset interest points. The historical log data between any two preset interest points may be obtained, and a corresponding sample log event sequence may be generated. The historical log data between any two preset interest points may refer to the historical log data generated when the terminal device transfers between the two preset interest points. Each sample log event sequence may correspond to a starting interest point and an ending interest point. The starting interest point refers to the interest point at which the sample log event sequence starts, and the ending interest point refers to the interest point at which the sample log event sequence ends. For example, Figure 1 As shown, assuming that the sample log event sequence is a sample log event sequence from position point A to position point B, the starting interest position point corresponding to the sample log event sequence is position A, and the ending interest position point is position point B; assuming that the sample log event sequence is a sample log event sequence from position point B to position point A, the starting interest position point corresponding to the sample log event sequence is position B, and the ending interest position point is position point A.

[0072] Therefore, the terminal device can generate multiple sample log event sequences between any two preset interest points based on historical log data, and use the sample log event sequences as sample data of the interest point inference model. The terminal interest point corresponding to the sample log event sequence can be used as the label corresponding to the sample data to train the interest point inference model.

[0073] The interest location point inference model to be trained is trained through multiple sample log event sequences generated by historical log data and the terminal interest locations corresponding to each sample log event sequence. It not only relies on the location data of the terminal device, but also makes the sample data used to train the interest location point inference model more comprehensive and accurately reflects the user's real behavior and intention, which can improve the accuracy of the interest location point inference model, thereby improving the accuracy of using the interest location point inference model to predict the interest location point that the terminal device is about to reach.

[0074] In some embodiments, after obtaining the historical log data of the terminal device, the terminal device may further perform the following steps: extracting historical location data from the historical log data of the terminal device; clustering the historical location data, and determining a plurality of preset interest locations based on the clustering results.

[0075] The historical log data includes historical location data, that is, the location data generated by the terminal device in the past period of time, which may include base station positioning data and GPS positioning data.

[0076] Optionally, the terminal device may cluster the historical location data in a manner that includes but is not limited to: density-based spatial clustering of applications with noise (DBSCAN) algorithm, maximum density clustering application (MDCA) algorithm, ordering points to identify the clustering structure (OPTICS) algorithm, etc.

[0077] The terminal device clusters the historical location data, and the clustering result includes multiple clusters. The center point of the cluster is the center of the location point frequently visited by the user. Therefore, the center point of the cluster can be used as the preset interest location point. Figure 3 As shown, Figure 3 is a schematic diagram of a clustering result disclosed in an embodiment of the present application. The multiple preset interest locations determined according to the clustering result include home D, company E, and shopping mall F.

[0078] Since historical location data can objectively record information about users' actual activities and can capture a wide range of users' activities, compared to directly obtaining preset interest points input by users, clustering historical location data to determine preset interest points can avoid interest points that users may ignore or miss, and discover potential interest points that users themselves are not aware of, which helps to comprehensively and accurately obtain preset interest points.

[0079] In some embodiments, the interest location point reasoning model may include but is not limited to a dual-tower model, such as a deep semantic matching model (Deep Structured Sematic Models, DSSM), a deep factorization machine (Deep Factorization Machines, DeepFM) model, a neural factorization machine (Neural Factorization Machines) model, etc.

[0080] Since the points of interest may be places that users frequently visit, such as home and company, and the users' daily commuting habits between home and company are similar, the log data generated by the terminal device between home and company is also similar. In an embodiment of the present application, a predicted log event sequence for input into the point of interest inference model can be generated based on the log data generated by the terminal device between the previous point of interest and the current point of interest, and the next point of interest that the terminal device will reach can be inferred.

[0081] The embodiment of the present application obtains the log data of the terminal device and predicts the next location point of interest of the terminal device, thereby reducing the power consumption of the terminal device caused by real-time positioning through GPS; and, based on the log data generated by the terminal device between the previous location point of interest and the current location point, it is possible to generate a predicted log event sequence in advance before the terminal device reaches the next location point of interest. The generated predicted log event sequence can accurately reflect the user's true behavior and intention, and determine the next location point of interest of the terminal device according to the input predicted log event sequence through the interest location point reasoning model, so that the next location point of interest to be reached by the user can be determined in advance when the user has not yet reached the location point of interest, which can improve the timeliness and accuracy of determining the location point of interest that the terminal device is about to reach.

[0082] Please read further Figure 4 , Figure 4 is a flow chart of another method for determining a point of interest location disclosed in an embodiment of the present application, which method can be applied to a terminal device; Figure 4 As shown, the method for determining the point of interest location may include the following steps:

[0083] 401. Obtain historical log data of the terminal device.

[0084] 402. Determine one or more complete log event sequences from a first interest point to a second interest point based on historical log data.

[0085] The complete log event sequence is a complete log event sequence generated during the transfer of the terminal device from the first interest point to the second interest point.

[0086] If the log event sequence generated during each transfer of the terminal device between the first interest point and the second interest point is the same, then there is a complete log event sequence corresponding to the first interest point and the second interest point.

[0087] Since the log event sequence generated each time the terminal device transfers between the first interest point and the second interest point may be different. For example, the terminal device took the subway for the first time, and a subway event occurred; the second time, the terminal device did not take the subway but drove, so no subway event occurred, and the terminal device may be connected to the car Bluetooth, so a Bluetooth connection event and a Bluetooth disconnection event occurred. Therefore, there may be multiple complete log event sequences corresponding to the first interest point and the second interest point.

[0088] The first interest point and the second interest point are any two preset interest points.

[0089] For example, assuming that the first interest point is a home and the second interest point is a company, please refer to Figure 5a , Figure 5a It is a schematic diagram of a complete log event sequence disclosed in an embodiment of the present application. The complete log event sequence is a log event sequence generated during the transfer of the terminal device between the two preset interest points of home and company. For example, the complete log event sequence from the first interest point (home) to the second interest point (company) may include: base station positioning event a——WiFi disconnection event b——elevator event c——walking event d——subway event e——music application usage event f——GPS positioning event g——social application usage event h——subway event i——work clock-in application usage event j——walking event k——elevator event l——WiFi connection event m——base station positioning event n. It should be noted that when base station positioning event a occurs, the terminal device determines that the location point of the terminal device is the first interest point (home) according to the base station positioning data; when base station positioning event n occurs, the terminal device determines that the location point of the terminal device is the second interest point (company) according to the base station positioning data.

[0090] 403. A log event sequence of a target duration is intercepted from each complete log event sequence to obtain a sample log event sequence corresponding to each complete log event sequence and having a termination interest point as a second interest point.

[0091] The terminal device captures a log event sequence of target duration from the complete log event sequence. The log event sequence of target duration may be captured starting from the first interest point in the complete log event sequence, or the log event sequence of target duration may be captured starting from an intermediate interest point between the first interest point and the second interest point in the complete log event sequence. There is no specific limitation.

[0092] For example, the terminal device is Figure 5a The log event sequence of the target duration obtained from the complete log event sequence shown can be as follows Figure 5b As shown, Figure 5b It is a schematic diagram of a sample log event sequence intercepted from a complete log event sequence disclosed in an embodiment of the present application. The sample log event sequence may be "base station positioning event a——WiFi disconnection event b——elevator event c——walking event d——subway event e——music application usage event f——GPS positioning event g——social application usage event h——subway event i". Alternatively, it may be "elevator event c——walking event d——subway event e——music application usage event f——GPS positioning event g——social application usage event h——subway event i——work clock-in application usage event j", without specific limitation.

[0093] Therefore, the sample log event sequence intercepted from the complete log event sequence can be used as the sample data of the interest point inference model, and the second interest point corresponding to the complete log event sequence can be used as the end interest point corresponding to the sample log event sequence, that is, the second interest point corresponding to the complete log event sequence can be used as the label corresponding to the sample log event sequence. Figure 5a The sample log event sequence intercepted from the complete log event sequence shown is base station positioning event a ~ subway event i, or elevator event c ~ work clock-in application usage event j. The terminal interest location points corresponding to the sample log event sequence are all the second interest location points (company) corresponding to the complete log time sequence, that is, the location point where the terminal device is located when the base station positioning event n occurs.

[0094] Optionally, the target duration may be a parameter input by the user, or may be positively correlated with the movement duration of the terminal device between the first interest point and the second interest point, and is not specifically limited. For example, the target duration may be 10 minutes or 20 minutes; the longer the movement duration of the terminal device between the first interest point and the second interest point, the longer the target duration.

[0095] In some embodiments, before extracting a log event sequence of target duration from each complete log event sequence and obtaining a sample log event sequence corresponding to each complete log event sequence and with the termination point of interest being the second point of interest, the terminal device may also perform the following steps: calculating the movement duration of the terminal device between the first point of interest and the second point of interest, and determining the target duration based on the movement duration and the duration factor.

[0096] The moving time is the time required for the terminal device to move between the first location of interest and the second location of interest. For example, if the time point at which the terminal device is at the first location of interest is 8 a.m. and the time point at which the terminal device is at the second location of interest is 9 a.m., the moving time is 1 hour.

[0097] Optionally, the target duration may be equal to the product of the moving duration and the duration factor; the duration factor may be 0.25 or 0.5, which is not specifically limited. Alternatively, the target duration may be equal to the result of dividing the moving duration and the duration factor; the duration factor may be 4 or 2, which is not specifically limited.

[0098] Through the above embodiments, log event sequences of target duration are intercepted from each complete log event sequence as sample log event sequences, which helps to enrich sample data used to train the interest location point inference model, thereby improving the generalization ability of the model, reducing the risk of overfitting, and improving the robustness of the model, thereby improving the accuracy of interest location point determination.

[0099] 404. Train the interest point inference model to be trained according to the multiple sample log event sequences and the termination interest point corresponding to each sample log event sequence.

[0100] The interest point inference model is trained based on multiple sample log event sequences and the terminal interest point corresponding to each sample log event sequence.

[0101] 405. In response to a positioning event occurring in the terminal device, calculate the distances between the current location point of the terminal device and a plurality of preset interest locations.

[0102] The preset interest points may include points input by the user, or may include points frequently visited by the user determined by the terminal device based on historical log data, without specific limitation.

[0103] In some embodiments, the terminal device determines preset interest locations by: extracting historical location data from historical log data of the terminal device; clustering the historical location data, and determining multiple preset interest locations based on the clustering results.

[0104] In some embodiments, the terminal device can determine that a positioning event has occurred in the terminal device when the terminal device is performing base station positioning or GPS positioning based on log data, and respond to the positioning event that has occurred in the terminal device. Optionally, the terminal device can perform base station positioning at a preset sampling interval, such as performing base station positioning every 5 minutes; the terminal device can perform GPS positioning when the user actively requests a positioning GPS service, when the user turns on the positioning service of the terminal device, or when the user opens a navigation application.

[0105] It can be seen that the embodiment of the present application calculates the distances between the current location of the terminal device and multiple preset interest locations respectively only when a positioning event occurs in the terminal device, so as to determine whether the terminal device has reached any preset interest location. Therefore, compared with the prior art of performing GPS positioning in real time or performing GPS positioning at short intervals, the embodiment of the present application combines low-power base station positioning with high-precision GPS positioning, and only when a positioning event occurs in the terminal device, calculates the distances between the current location of the terminal device and multiple preset interest locations respectively, so as to determine whether the terminal device has reached any preset interest location, and infers the next interest location of the terminal device in advance when the terminal device has not reached any preset interest location, thereby reducing the power consumption of the terminal device caused by actively requesting GPS positioning, and can improve the timeliness and accuracy of determining the interest location that the terminal device is about to reach.

[0106] 406. If it is determined that the current location point does not belong to any preset interest points based on the distances between the current location point and the multiple preset interest points, log data generated by the terminal device between the previous interest point and the current location point is obtained.

[0107] The terminal device can calculate the distances between the current location and multiple preset interest locations. If there is a preset interest location whose distance from the current location is less than a distance threshold, it can be determined that the current location of the terminal device belongs to the preset interest location, that is, the terminal device has reached the preset interest location, and there is no need to execute steps 406 to 408.

[0108] If the distance between the current location point and each preset point of interest is greater than the distance threshold, it can be determined that the current location point of the terminal device does not belong to any preset point of interest, that is, the terminal device has not reached any preset point of interest. Therefore, the log data generated by the terminal device between the previous point of interest and the current point of interest can be obtained, and the subsequent steps of determining the next point of interest of the terminal device can be executed.

[0109] 407. Generate a predicted log event sequence based on the log data.

[0110] The embodiment of the present application adopts a real-time perception and recommendation strategy to analyze and process the log event sequence of the terminal device in real time. By matching and comparing with the real-time location data, the points of interest that the terminal device is about to reach can be updated in time, providing users with real-time, personalized points of interest recommendation services. Compared with the traditional method of determining points of interest, the embodiment of the present application has a stronger real-time perception and recommendation capability for points of interest, and users can obtain recommendations for points of interest related to their current needs in a timely manner.

[0111] 408. Textually process the predicted log event sequence to obtain a predicted text sequence, and input the predicted text sequence into the trained interest point inference model to determine the next interest point of the terminal device according to the predicted text sequence through the interest point inference model.

[0112] Textual processing of the predicted log event sequence, that is, describing the predicted log event sequence with text data, helps to simplify the process of preprocessing the data input to the interest location point reasoning model, and converts the data input to the interest location point reasoning model into text data, which helps to improve the efficiency of data analysis and modeling.

[0113] For example, assuming that the predicted log event sequence is "base station positioning event - WiFi disconnection event - elevator event - walking event - subway event - music application usage event - GPS positioning event - social application usage event - subway event - work clocking-in application usage event - walking event - elevator event - WiFi connection event - base station positioning event", the predicted text sequence obtained after textual processing of the predicted log event sequence can be: [base station positioning, WiFi disconnection, elevator, walking, subway, music application, GPS positioning, social application, subway, work clocking-in application, walking, elevator, WiFi connection, base station positioning].

[0114] It can be seen that the method for determining points of interest in the embodiment of the present application can realize the prediction of complex user behavior sequences based on the log event sequence of the terminal device, more comprehensively and accurately perceive the points of interest that the user is about to reach, and reduce the power consumption of the mobile phone caused by actively requesting GPS positioning. Therefore, points of interest related to the user's current needs and consistent with behavioral habits can be recommended to the user, which can greatly improve the user experience and user satisfaction.

[0115] In some embodiments, after determining the next point of interest of the terminal device according to the predicted text sequence through the point of interest inference model, the terminal device can calculate the distance between the current point of interest and the next point of interest, and calculate the speed according to the distance and movement time between the previous point of interest and the current point of interest of the terminal device, and calculate the time required for the terminal device to move from the current point of interest to the next point of interest according to the distance and the speed, thereby predicting how long it will take for the user to reach the next point of interest. Alternatively, the historical location data of the terminal device can be input into the time prediction model, and the time prediction model can be used to predict the time required for the terminal device to move from the current point of interest to the next point of interest.

[0116] The embodiment of the present application determines one or more complete log event sequences from a first interest point to a second interest point of a terminal device based on historical log data, and extracts a log event sequence of a target duration from the complete log event sequence as a sample log event sequence corresponding to the complete log event and with the termination interest point being the second interest point, which helps to enrich the sample data used to train the interest point inference model, thereby improving the generalization ability of the model and reducing the risk of overfitting, thereby improving the robustness of the model; the interest point inference model to be trained is trained based on multiple sample log event sequences and the termination interest points corresponding to each sample log event sequence; in response to a positioning event occurring at the terminal device, the current position of the terminal device is calculated only when a positioning event occurs at the terminal device. The distance between the set point and multiple preset interest points is respectively determined, so as to determine whether the terminal device has reached any preset interest point, and obtain the log data of the terminal device when the terminal device has not reached any preset interest point, thereby reducing the power consumption of the terminal device caused by real-time positioning through GPS, and generating a predicted log event sequence in advance when the terminal device has not reached the next interest point. The generated predicted log event sequence can accurately reflect the user's real behavior and intention, and determine the next interest point of the terminal device according to the input predicted log event sequence through the interest point reasoning model, so that the next interest point to be reached by the user can be determined in advance when the user has not reached the interest point, which can improve the timeliness and accuracy of determining the interest point that the terminal device is about to reach.

[0117] Please read further Figure 6 , Figure 6 is a flow chart of a method for training an interest point inference model disclosed in an embodiment of the present application, which method can be applied to a terminal device; Figure 6 As shown, the training method of the interest location point inference model may include the following steps:

[0118] 601. Obtain historical log data of the terminal device.

[0119] The historical log data may be log data generated by the terminal device in the past period of time. Optionally, the log data may include at least one of wireless connection data, behavior recognition data, location data, and application usage data.

[0120] 602. Generate multiple sample log event sequences between any two preset interest locations based on historical log data.

[0121] In some embodiments, the terminal device generates a plurality of sample log event sequences between any two preset interest points according to the historical log data, which may include the following steps:

[0122] According to historical log data, one or more complete log event sequences from a first interest point to a second interest point are determined; the first interest point and the second interest point are any two preset interest points; a log event sequence of a target duration is intercepted from each complete log event sequence to obtain a sample log event sequence corresponding to each complete log event sequence, and the termination interest point is the second interest point.

[0123] In some embodiments, before the terminal device intercepts a log event sequence of target duration from each complete log event sequence and obtains a sample log event sequence with the termination point of interest being the second point of interest, it may also perform the following steps: calculate the movement duration of the terminal device between the first point of interest and the second point of interest, and determine the target duration based on the movement duration and the duration factor.

[0124] In some embodiments, after obtaining the historical log data of the terminal device, the terminal device may further perform the following steps: extracting historical location data from the historical log data of the terminal device; clustering the historical location data, and determining a plurality of preset interest locations based on the clustering results.

[0125] 603. Obtain a user portrait corresponding to the terminal device.

[0126] User portraits can be automatically generated by terminal devices based on user behavior data and personal information. User portraits can reflect user behavior trajectories, activity habits, interest preferences, etc.

[0127] When the above embodiments of the present application are applied to specific products or technologies, the subject's permission or consent must be obtained, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards of relevant countries and regions.

[0128] The terminal device can train the interest point inference model to be trained based on the user portrait, multiple sample log event sequences, the terminal interest point corresponding to each sample log event sequence, and multiple preset interest points.

[0129] Optionally, the terminal device can perform feature extraction on the user portrait, the first sample log event sequence, and multiple preset interest points, respectively, to obtain portrait features corresponding to the user portrait, log features corresponding to the first sample log event sequence, and interest point features corresponding to the multiple preset interest points; and perform weighted summation of the portrait features and the log features to obtain user features; determine the predicted interest points based on the user features and the interest point features corresponding to the multiple preset interest points; calculate the training loss based on the predicted interest points and the termination interest points corresponding to the first sample log event sequence, and adjust the parameters of the interest point inference model to be trained based on the training loss.

[0130] Since each user has personalized characteristics, integrating log event sequences and user portraits as user features can more accurately describe the user's complex behavior patterns and more accurately perceive points of interest. Compared with methods that rely heavily on location information, the feature extraction of the embodiment of the present application is more comprehensive and can better reflect the user's real behavior and intentions, thereby improving the accuracy of the interest point reasoning model, thereby improving the accuracy of using the interest point reasoning model to predict the interest point that the terminal device is about to reach.

[0131] 604. Perform feature extraction on the user portrait, the first sample log event sequence, and multiple preset interest locations, respectively, to obtain portrait features corresponding to the user portrait, log features corresponding to the first sample log event sequence, and interest point features corresponding to the multiple preset interest locations, respectively.

[0132] The first sample log event sequence is any sample log event sequence.

[0133] Optionally, the feature extraction method may include but is not limited to a convolutional neural network (CNN), a recurrent neural network (RNN), and a transformer model supporting a self-attention mechanism.

[0134] In some embodiments, before the terminal device extracts features from the user portrait, the first sample log event sequence, and multiple preset interest points, the terminal device may also perform text processing on the sample log event sequence. Therefore, the data types of the sample log event sequence and the user portrait are both text types. For example, the sample log event sequence may be [base station positioning, WiFi disconnection, elevator, walking, subway, music application, GPS positioning, social application, subway], and the user portrait may be "likes sports, likes to share news on social media, and often listens to music"; the data type of the preset interest point is a numerical type, for example, the preset interest point may be a latitude and longitude value.

[0135] 605. Concatenate the portrait features and the log features to obtain user features.

[0136] By concatenating portrait features and log features, the user features obtained by concatenating multi-dimensional features can increase the interest location point reasoning model, which helps the model to better understand the complexity and diversity of the data; and, concatenating multi-dimensional features together can capture the association and interaction between different features, which helps to provide more comprehensive information for model training, so that the model can more accurately learn the relationship between various data and improve the generalization and learning ability of the interest location point reasoning model.

[0137] The terminal device may determine the predicted interest point according to the user characteristics and the interest point characteristics corresponding to the plurality of preset interest points. Optionally, the terminal device may calculate the correlation or similarity score between the user characteristics and the characteristics of each interest point, such as cosine similarity, Euclidean distance, etc.

[0138] 606. Calculate the cosine similarity between the user feature and each point of interest feature, and determine the point of interest feature with the largest cosine similarity.

[0139] Cosine similarity is an indicator to measure the similarity between two feature vectors, which can refer to the cosine value of the angle between two feature vectors in the vector space. The value range of cosine similarity can be between 0 and 1, the closer to 1, the more similar, and 0 means no similarity. Cosine similarity can be used to calculate the similarity between text type data, and is applied to tasks such as recommendation systems, information retrieval, and text classification.

[0140] 607. Determine the preset interest point corresponding to the interest point feature with the largest cosine similarity as the predicted interest point corresponding to the first sample log event sequence.

[0141] For example, if the cosine similarities between the user feature and each interest point feature are 0.25, 0.65, 0.05, and 0.05 respectively, the interest point feature with a cosine similarity of 0.65 with the user feature is taken as the interest point feature with the largest cosine similarity, and the preset interest location point corresponding to the interest point feature with the largest cosine similarity is determined as the predicted interest location point corresponding to the first sample log event sequence.

[0142] 608. Calculate the training loss based on the predicted interest location and the termination interest location corresponding to the first sample log event sequence, and adjust the parameters of the interest location inference model to be trained based on the training loss.

[0143] The predicted interest point corresponding to the first sample log event sequence is the predicted value output by the interest point inference model to be trained, and the terminal interest point corresponding to the first sample log event sequence is the true label. The training loss can be used to measure the difference between the predicted value output by the interest point inference model to be trained based on the sample data and the true label, and is used to guide the adjustment of the model parameters so that the predicted value is closer to the true label.

[0144] The predicted points of interest and the final points of interest are both of numerical type, for example, latitude and longitude values.

[0145] The method for calculating the training loss based on the predicted interest point and the termination interest point corresponding to the first sample log event sequence may include but is not limited to methods such as calculating the mean squared error (MSE), calculating the cross-entropy loss (Cross-Entropy Loss), and calculating the logarithmic loss (Log Loss).

[0146] Methods for adjusting parameters of the interest location point inference model to be trained according to the training loss may include but are not limited to gradient descent (GD), stochastic gradient descent (SGD) and adaptive learning rate algorithms (Adaptive Learning Rate Algorithms).

[0147] Optionally, the interest location point inference model can also integrate data such as voice and pictures to more comprehensively describe the user's behavior and environment and improve the accuracy of interest location point perception.

[0148] See also Figure 7 , Figure 7 The structure diagram of an interest point reasoning model disclosed in the embodiment of the present application is shown in FIG. The interest point reasoning model includes an input layer, a representation layer and a matching layer.

[0149] The input layer is used to input user portraits, multiple sample log event sequences, the end interest points corresponding to each sample log event sequence, and multiple preset interest points.

[0150] The representation layer is used to extract features of the user portrait, the first sample log event sequence, and multiple preset interest locations through a neural network, so as to obtain portrait features corresponding to the user portrait, log features corresponding to the first sample log event sequence, and interest point features corresponding to the multiple preset interest locations, and to concatenate the portrait features and log features to obtain user features.

[0151] The matching layer is used to calculate the cosine similarity between user features and features of each point of interest.

[0152] See also Figure 8 , Figure 8 It is a schematic diagram of the reasoning process of an interest location point reasoning model disclosed in an embodiment of the present application.

[0153] The previous location point of interest of the terminal device is location point G, and the current location point of the terminal device is location point H. The predicted log event sequence K generated by the terminal device between location point G and location point H is input into the trained interest location point inference model, so that the interest location point inference model is used to infer the scores corresponding to each preset interest location point according to the predicted log event sequence K, user portrait, and multiple preset interest locations, and the preset interest location point with the highest score is determined as the next interest location point of the terminal device.

[0154] In the embodiment of the present application, the user portrait and the log event sequence are integrated to train the interest point inference model, which can more accurately describe the user's complex behavior pattern and more accurately perceive the interest point. The portrait features extracted from the user portrait and the log features extracted from the first sample log event sequence are concatenated to obtain the user features, and the interest point features are extracted from multiple preset interest points, realizing multi-dimensional feature extraction. Compared with the method with a high degree of dependence on location information, the feature extraction of the embodiment of the present application is more comprehensive and can better reflect the user's real behavior and intention; and based on the cosine similarity between the user features and the features of each interest point, the predicted interest point corresponding to the first sample log event sequence is determined, and the training loss is calculated according to the predicted interest point and the terminal interest point corresponding to the first sample log event sequence. The model is adjusted and optimized. Through the training method of the interest point inference model, the accuracy of the interest point inference model can be improved, thereby improving the accuracy of using the interest point inference model to predict the interest point that the terminal device is about to reach.

[0155] See also Fig. 9 , Fig. 9 is a schematic diagram of a structure of a device for determining a point of interest disclosed in an embodiment of the present application. The device can be applied to a terminal device, such as Fig. 9 As shown, the device 900 for determining a point of interest location may include: a first acquisition module 910 , a first generation module 920 and a reasoning module 930 .

[0156] A first acquisition module 910 is used to acquire log data generated by the terminal device between the last interest point and the current position point;

[0157] A first generating module 920, configured to generate a predicted log event sequence according to log data;

[0158] The inference module 930 is also used to input the predicted log event sequence into the trained interest location point inference model, so as to determine the next interest location point of the terminal device according to the predicted log event sequence through the interest location point inference model; the interest location point inference model is trained based on multiple sample log event sequences and the terminal interest location points corresponding to each sample log event sequence.

[0159] In one embodiment, the first acquisition module 910 is also used to calculate the distances between the current location point of the terminal device and multiple preset interest points in response to a positioning event occurring in the terminal device; if it is determined that the current location point does not belong to any preset interest points based on the distances between the multiple preset interest points, then the log data generated by the terminal device between the previous interest point and the current location point is obtained.

[0160] In one embodiment, the log data includes at least one of wireless connection data, behavior recognition data, location data, and application usage data.

[0161] In one embodiment, the reasoning module 930 is also used to perform text processing on the predicted log event sequence to obtain a predicted text sequence, and input the predicted text sequence into the trained interest location point reasoning model to determine the next interest location point of the terminal device according to the predicted text sequence through the interest location point reasoning model.

[0162] In one embodiment, the device 900 for determining a point of interest location further includes a second training module;

[0163] The second training module is used to obtain the historical log data of the terminal device; generate multiple sample log event sequences between any two preset interest points based on the historical log data; and train the interest point inference model to be trained based on the multiple sample log event sequences and the terminal interest points corresponding to each sample log event sequence.

[0164] In one embodiment, the second training module is also used to determine one or more complete log event sequences from a first interest point to a second interest point based on historical log data; the first interest point and the second interest point are any two preset interest points; a log event sequence of a target duration is intercepted from each complete log event sequence to obtain a sample log event sequence corresponding to each complete log event sequence, and the termination interest point is the second interest point.

[0165] In one embodiment, the second training module is also used to calculate the movement duration of the terminal device between the first point of interest and the second point of interest before intercepting the log event sequence of the target duration from each complete log event sequence to obtain a sample log event sequence with the termination point of interest being the second point of interest, and determine the target duration based on the movement duration and the duration factor.

[0166] In one embodiment, the second training module is further used to extract historical location data from the historical log data of the terminal device; cluster the historical location data, and determine a plurality of preset interest locations according to the clustering results.

[0167] In one embodiment, the second training module is also used to obtain a user portrait corresponding to the terminal device before training the interest location inference model to be trained based on multiple sample log event sequences and the terminal interest location corresponding to each sample log event sequence; and train the interest location inference model to be trained based on the user portrait, multiple sample log event sequences, the terminal interest location corresponding to each sample log event sequence, and multiple preset interest locations.

[0168] In one embodiment, the second training module is also used to extract features from the user portrait, the first sample log event sequence, and multiple preset interest points, respectively, to obtain portrait features corresponding to the user portrait, log features corresponding to the first sample log event sequence, and interest point features corresponding to the multiple preset interest points; the first sample log event sequence is any sample log event sequence; the portrait features and the log features are concatenated to obtain user features; the predicted interest points are determined based on the user features and the interest point features corresponding to the multiple preset interest points; the training loss is calculated based on the predicted interest points and the termination interest points corresponding to the first sample log event sequence, and the parameters of the interest point inference model to be trained are adjusted based on the training loss.

[0169] In one embodiment, the second training module is also used to calculate the cosine similarity between the user features and the features of each point of interest, and determine the point of interest feature with the largest cosine similarity; and determine the preset interest location point corresponding to the point of interest feature with the largest cosine similarity as the predicted interest location point corresponding to the first sample log event sequence.

[0170] The embodiment of the present application obtains the log data of the terminal device and predicts the next location point of interest of the terminal device, thereby reducing the power consumption of the terminal device caused by real-time positioning through GPS; and, based on the log data generated by the terminal device between the previous location point of interest and the current location point, it is possible to generate a predicted log event sequence in advance before the terminal device reaches the next location point of interest. The generated predicted log event sequence can accurately reflect the user's true behavior and intention, and determine the next location point of interest of the terminal device according to the input predicted log event sequence through the interest location point reasoning model, so that the next location point of interest to be reached by the user can be determined in advance when the user has not yet reached the location point of interest, which can improve the timeliness and accuracy of determining the location point of interest that the terminal device is about to reach.

[0171] See also Fig.10 , Fig.10 is a schematic diagram of a training device for an interest point inference model disclosed in an embodiment of the present application. The device can be applied to a terminal device, such as Fig.10 As shown, the training device 1000 for the interest location point reasoning model may include: a second acquisition module 1010, a second generation module 1020 and a first training module 1030;

[0172] The second acquisition module 1010 is used to acquire historical log data of the terminal device;

[0173] The second generating module 1020 is used to generate a plurality of sample log event sequences between any two preset interest locations according to the historical log data;

[0174] The first training module 1030 is used to train the interest point inference model to be trained according to multiple sample log event sequences and the terminal interest point corresponding to each sample log event sequence.

[0175] In one embodiment, the first training module 1030 is also used to obtain a user portrait corresponding to the terminal device before training the interest location inference model to be trained based on multiple sample log event sequences and the terminal interest location corresponding to each sample log event sequence; and train the interest location inference model to be trained based on the user portrait, multiple sample log event sequences, the terminal interest location corresponding to each sample log event sequence, and multiple preset interest locations.

[0176] In one embodiment, the first training module 1030 is also used to extract features from the user portrait, the first sample log event sequence, and multiple preset interest points, respectively, to obtain portrait features corresponding to the user portrait, log features corresponding to the first sample log event sequence, and interest point features corresponding to the multiple preset interest points; the first sample log event sequence is any sample log event sequence; the portrait features and the log features are concatenated to obtain user features; the predicted interest points are determined based on the user features and the interest point features corresponding to the multiple preset interest points; the training loss is calculated based on the predicted interest points and the termination interest points corresponding to the first sample log event sequence, and the parameters of the interest point inference model to be trained are adjusted based on the training loss.

[0177] In one embodiment, the first training module 1030 is also used to calculate the cosine similarity between the user features and each point of interest feature, and determine the point of interest feature with the largest cosine similarity; the preset point of interest location corresponding to the point of interest feature with the largest cosine similarity is determined as the predicted point of interest location corresponding to the first sample log event sequence.

[0178] The embodiment of the present application trains the interest location point inference model to be trained through multiple sample log event sequences generated by historical log data and the terminal interest location points corresponding to each sample log event sequence, and does not rely solely on the location data of the terminal device, so that the sample data used to train the interest location point inference model is more comprehensive and accurately reflects the user's real behavior and intentions, which can improve the accuracy of the interest location point inference model, thereby improving the accuracy of using the interest location point inference model to predict the interest location point that the terminal device is about to reach.

[0179] See also Fig.11 , Fig.11 Schematic diagram of the structure of a terminal device disclosed in the embodiment of the present application. Fig.11 As shown, the terminal device 1100 may include:

[0180] A memory 1110 storing executable program codes;

[0181] a processor 1120 coupled to the memory 1110;

[0182] The processor 1120 calls the executable program code stored in the memory 1110 to execute any one of the methods for determining a point of interest disclosed in the embodiments of the present application.

[0183] An embodiment of the present application discloses a computer-readable storage medium storing a computer program, wherein when the computer program is executed by the processor, the processor implements any one of the methods for determining a location of interest disclosed in the embodiment of the present application.

[0184] It should be understood that the "one embodiment" or "an embodiment" mentioned throughout the specification means that the specific features, structures or characteristics related to the embodiment are included in at least one embodiment of the present application. Therefore, "in one embodiment" or "in an embodiment" appearing throughout the specification does not necessarily refer to the same embodiment. In addition, these specific features, structures or characteristics can be combined in one or more embodiments in any suitable manner. Those skilled in the art should also be aware that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily required for the present application.

[0185] In the various embodiments of the present application, it should be understood that the size of the serial numbers of the above-mentioned processes does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0186] The units described above as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0187] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0188] If the above-mentioned integrated unit 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-accessible memory. Based on this understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product, which is stored in a memory and includes several requests for a computer device (which can be a personal computer, a server or a network device, etc., specifically a processor in a computer device) to perform some or all of the steps of the above-mentioned methods of various embodiments of the present application.

[0189] Those skilled in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium, and the storage medium includes a read-only memory (ROM), a random access memory (RAM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electronically erasable rewritable read-only memory (EEPROM), a compact disc (CD-ROM) or other optical disc storage, magnetic disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0190] The above is a detailed introduction to a method, device, terminal device and storage medium for determining a point of interest disclosed in an embodiment of the present application. This article uses specific examples to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method and core idea of ​​the present application. At the same time, for those of ordinary skill in the art, according to the idea of ​​the present application, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A method for determining a point of interest, characterized in that: Applied to a terminal device, the method comprises: Obtaining log data generated by the terminal device between the last point of interest and the current point of interest; generating a predicted log event sequence according to the log data; The predicted log event sequence is input into the trained interest location point inference model to determine the next interest location point of the terminal device according to the predicted log event sequence through the interest location point inference model; the interest location point inference model is trained based on multiple sample log event sequences and the terminal interest location points corresponding to each of the sample log event sequences.

2. The method according to claim 1, characterized in that The obtaining of log data generated by the terminal device between the last interest point and the current position point includes: In response to a positioning event occurring at the terminal device, calculating the distances between a current location point of the terminal device and a plurality of preset interest locations; If it is determined based on the distances between the current location point and the plurality of preset interest points that the current location point does not belong to any of the preset interest points, log data generated by the terminal device between the last interest point and the current location point is obtained.

3. The method according to claim 1, characterized in that The log data includes at least one of wireless connection data, behavior recognition data, location data, and application usage data.

4. The method according to claim 1, characterized in that: The step of inputting the predicted log event sequence into the trained interest point inference model to determine the next interest point of the terminal device according to the predicted log event sequence through the interest point inference model includes: The predicted log event sequence is processed into text form to obtain a predicted text sequence, and the predicted text sequence is input into a trained interest point inference model to determine the next interest point of the terminal device according to the predicted text sequence through the interest point inference model.

5. The method according to any one of claims 1 to 4, characterized in that: Before acquiring the log data generated by the terminal device between the last interest point and the current position point, the method further includes: Obtaining historical log data of the terminal device; Generating a plurality of sample log event sequences between any two preset interest locations according to the historical log data; The interest point inference model to be trained is trained according to a plurality of sample log event sequences and the termination interest points corresponding to each of the sample log event sequences.

6. The method according to claim 5, characterized in that The step of generating a plurality of sample log event sequences between any two preset interest points according to the historical log data includes: Determine one or more complete log event sequences from a first point of interest to a second point of interest based on the historical log data; the first point of interest and the second point of interest are any two preset points of interest; A log event sequence of target duration is intercepted from each of the complete log event sequences to obtain a sample log event sequence corresponding to each of the complete log event sequences and having a termination interest point as the second interest point.

7. The method according to claim 6, characterized in that Before intercepting a log event sequence of target duration from each of the complete log event sequences to obtain a sample log event sequence corresponding to each of the complete log event sequences and having a termination point of interest being the second point of interest, the method further includes: The moving duration of the terminal device between the first interest point and the second interest point is calculated, and a target duration is determined according to the moving duration and a duration factor.

8. The method according to claim 5, characterized in that After acquiring the historical log data of the terminal device, the method further includes: extracting historical location data from historical log data of the terminal device; The historical location data is clustered, and a plurality of preset interest locations are determined according to the clustering results.

9. The method according to claim 5, characterized in that Before training the interest point inference model to be trained based on the plurality of sample log event sequences and the termination interest points corresponding to each of the sample log event sequences, the method further includes: Obtaining a user portrait corresponding to the terminal device; The training of the interest point inference model to be trained according to the plurality of sample log event sequences and the termination interest points corresponding to each of the sample log event sequences includes: The interest point inference model to be trained is trained according to the user portrait, multiple sample log event sequences, the termination interest points corresponding to each of the sample log event sequences, and multiple preset interest points.

10. The method according to claim 9, characterized in that The training of the interest point inference model to be trained according to the user portrait, the plurality of sample log event sequences, the termination interest points corresponding to each of the sample log event sequences, and the plurality of preset interest points includes: Performing feature extraction on the user portrait, the first sample log event sequence, and the plurality of preset interest points, respectively, to obtain a portrait feature corresponding to the user portrait, a log feature corresponding to the first sample log event sequence, and interest point features corresponding to the plurality of preset interest points, respectively; the first sample log event sequence is any of the sample log event sequences; Concatenate the portrait feature and the log feature to obtain user features; Determining a predicted point of interest based on the user characteristics and the point of interest characteristics respectively corresponding to the plurality of preset points of interest; A training loss is calculated based on the predicted interest point and the termination interest point corresponding to the first sample log event sequence, and parameters of the interest point inference model to be trained are adjusted based on the training loss.

11. The method according to claim 10, characterized in that The step of determining the predicted interest point according to the user feature and the interest point features respectively corresponding to the plurality of preset interest points comprises: Calculating the cosine similarity between the user feature and each of the interest point features, and determining the interest point feature with the largest cosine similarity; The preset interest point corresponding to the interest point feature with the largest cosine similarity is determined as the predicted interest point corresponding to the first sample log event sequence.

12. A method for training an interest point inference model, characterized in that: The method comprises: Obtain historical log data of terminal devices; Generating a plurality of sample log event sequences between any two preset interest locations according to the historical log data; The interest point inference model to be trained is trained according to a plurality of sample log event sequences and the termination interest points corresponding to each of the sample log event sequences.

13. The method according to claim 12, characterized in that Before training the interest point inference model to be trained based on the plurality of sample log event sequences and the termination interest points corresponding to each of the sample log event sequences, the method further includes: Obtaining a user portrait corresponding to the terminal device; The training of the interest point inference model to be trained according to the plurality of sample log event sequences and the termination interest points corresponding to each of the sample log event sequences includes: The interest point inference model to be trained is trained according to the user portrait, multiple sample log event sequences, the termination interest points corresponding to each of the sample log event sequences, and multiple preset interest points.

14. The method according to claim 13, characterized in that The training of the interest point inference model to be trained according to the user portrait, the plurality of sample log event sequences, the termination interest points corresponding to each of the sample log event sequences, and the plurality of preset interest points includes: Performing feature extraction on the user portrait, the first sample log event sequence, and the plurality of preset interest points, respectively, to obtain a portrait feature corresponding to the user portrait, a log feature corresponding to the first sample log event sequence, and interest point features corresponding to the plurality of preset interest points, respectively; the first sample log event sequence is any of the sample log event sequences; Concatenate the portrait feature and the log feature to obtain user features; Determining a predicted point of interest based on the user characteristics and the point of interest characteristics respectively corresponding to the plurality of preset points of interest; A training loss is calculated based on the predicted interest point and the termination interest point corresponding to the first sample log event sequence, and parameters of the interest point inference model to be trained are adjusted based on the training loss.

15. The method according to claim 14, characterized in that The step of determining the predicted interest point according to the user feature and the interest point features respectively corresponding to the plurality of preset interest points comprises: Calculating the cosine similarity between the user feature and each of the interest point features, and determining the interest point feature with the largest cosine similarity; The preset interest point corresponding to the interest point feature with the largest cosine similarity is determined as the predicted interest point corresponding to the first sample log event sequence.

16. A device for determining a point of interest, characterized in that: Applied to a terminal device, the device comprises: A first acquisition module, used to acquire log data generated by the terminal device between the last interested location point and the current location point; A first generating module, used to generate a predicted log event sequence according to the log data; An inference module is used to input the predicted log event sequence into a trained interest point inference model, so as to determine the next interest point of the terminal device according to the predicted log event sequence through the interest point inference model; the interest point inference model is trained based on multiple sample log event sequences and the terminal interest point corresponding to each of the sample log event sequences.

17. A training device for an interest point inference model, characterized in that: The device comprises: The second acquisition module is used to acquire historical log data of the terminal device; A second generating module is used to generate a plurality of sample log event sequences between any two preset interest locations according to the historical log data; The first training module is used to train the interest point inference model to be trained according to a plurality of sample log event sequences and the terminal interest point corresponding to each sample log event sequence.

18. A terminal device, characterized in that: The method comprises a memory and a processor, wherein a computer program is stored in the memory, and when the computer program is executed by the processor, the processor implements the method according to any one of claims 1 to 11 or 12 to 15.

19. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 11 or 12 to 15 is implemented.

Images