Point of interest state determination method, apparatus, and electronic device

Abstract

Embodiments of the present disclosure disclose a point of interest state determination method and device and electronic equipment. The method comprises: obtaining first feature data of a target point of interest in a first preset historical time period; inputting the first feature data into a pre-trained point of interest state determination model to predict a state probability distribution of the target point of interest in the first preset historical time period; comparing a state probability corresponding to a preset time point in the first preset historical time period in the state probability distribution with a preset probability threshold to determine a state of the target point of interest at the preset time point. The technical solution can quickly predict the state of the target point of interest, thereby notifying the user in time, and effectively improving the service quality of the Internet platform, protecting the user experience, and avoiding the loss of user time and distance.

Description

Technical Field

[0001] This disclosure relates to the field of geographic information technology, specifically to a method, apparatus, and electronic device for determining the status of points of interest. Background Technology

[0002] With the development of internet and data technology, more and more users are searching for and querying locations through internet platforms. However, in practical applications, the rate of change in the status of a location is relatively high, and it is difficult to notify every user of these changes in a timely manner. For example, if a restaurant closes, relocates, demolishes, or changes its name, even if its members are notified, users who are not members but want to dine there may not be informed in a timely manner. This causes great inconvenience to users' searches and may even result in time and travel losses for those who want to visit the restaurant. To solve the problem of not being able to notify users of location status changes in a timely manner, existing technologies mine location status changes. However, the mining speed of existing technologies is low, making it difficult to improve the service quality of internet platforms and guarantee the user experience. Summary of the Invention

[0003] This disclosure provides a method, apparatus, electronic device, and computer-readable storage medium for determining the state of a point of interest.

[0004] Firstly, this disclosure provides a method for determining the state of a point of interest.

[0005] Specifically, the method for determining the state of the point of interest includes:

[0006] Acquire the first feature data of the target point of interest within a first preset historical time period;

[0007] The first feature data is input into a pre-trained interest point state determination model to predict the state probability distribution of the target interest point within the first preset historical time period.

[0008] The state probability corresponding to the preset time point in the state probability distribution within the first preset historical time period is compared with a preset probability threshold to determine the state of the target point of interest at the preset time point.

[0009] In conjunction with the first aspect, in a first implementation of the first aspect of this disclosure, the method further includes:

[0010] The second feature data of the sample interest point is obtained, which is determined to be the expiration state time of the expiration state and the sample interest point is within a second preset historical time period before the expiration state time.

[0011] Based on the expired state time and the second feature data, the time state sequence of the sample interest points is estimated, and the estimated time state sequence of the sample interest points is obtained.

[0012] An initial interest point state determination model is determined, and the initial interest point state determination model is trained based on the second feature data and the estimated time state sequence to obtain the interest point state determination model.

[0013] In conjunction with the first aspect and the first implementation of the first aspect, in the second implementation of the first aspect of this disclosure, the step of estimating the time state sequence of sample interest points based on the expired state time and the second feature data includes:

[0014] Calculate the cumulative feature data distribution of the sample interest points within the second preset historical time period;

[0015] The time point corresponding to when the accumulated feature data is the first set value is obtained;

[0016] Set the probability that the sample point of interest is in an expired state within the time period before the time point mentioned above to a first preset value.

[0017] Set the probability that the sample point of interest corresponding to the expiration time is in an expired state to a second preset value;

[0018] For the third preset historical time period between the time point corresponding to the first preset value of the accumulated feature data and the time of the expired state, interpolation is performed using the first preset value and the second preset value as endpoint probability values ​​to obtain the probability that the sample point of interest is in an expired state within the third preset historical time period. This results in the probability that the sample point of interest is in an expired state within the third preset historical time period showing a monotonically increasing trend, thereby obtaining the probability sequence of the state probability value of the sample point of interest as a function of time, i.e., the estimated time state sequence of the sample point of interest.

[0019] Combining the first aspect, the first implementation of the first aspect, and the second implementation of the first aspect, in the third implementation of the first aspect of this disclosure, the step of training the initial interest point state determination model based on the second feature data and the estimated time state sequence includes:

[0020] Obtain a training data set, wherein the training data set includes second feature training data obtained by sliding segmentation of the second feature data and the estimated time state sequence, and the estimated time state training sequence corresponding to the second feature training data;

[0021] Feature extraction is performed on the second feature training data to obtain the third feature training data corresponding to the second feature training data.

[0022] Using the second feature training data and the third feature training data as input, and the corresponding estimated time state training sequence as output, the initial interest point state determination model is trained to obtain the interest point state determination model.

[0023] In conjunction with the first aspect, the first implementation of the first aspect, the second implementation of the first aspect, and the third implementation of the first aspect, in the fourth implementation of the first aspect of this disclosure, the step of extracting features from the second feature training data to obtain third feature training data corresponding to the second feature training data includes:

[0024] Extract local features from the training data for the second feature;

[0025] Extract the global features from the training data for the second feature;

[0026] The local and global features of the second feature training data are combined to form the third feature training data corresponding to the second feature training data.

[0027] Secondly, this disclosure provides a device for determining the state of a point of interest.

[0028] Specifically, the point of interest state determination device includes:

[0029] The acquisition module is configured to acquire the first feature data of the target point of interest within a first preset historical time period;

[0030] The prediction module is configured to input the first feature data into a pre-trained interest point state determination model to predict the state probability distribution of the target interest point within the first preset historical time period.

[0031] The determination module is configured to compare the state probability corresponding to the preset time point in the state probability distribution within the first preset historical time period with a preset probability threshold to determine the state of the target point of interest at the preset time point.

[0032] In conjunction with the second aspect, in a first implementation of the second aspect of this disclosure, the apparatus further includes a training module, wherein the training module is configured to:

[0033] The second feature data of the sample interest point is obtained, which is determined to be the expiration state time of the expiration state and the sample interest point is within a second preset historical time period before the expiration state time.

[0034] Based on the expired state time and the second feature data, the time state sequence of the sample interest points is estimated, and the estimated time state sequence of the sample interest points is obtained.

[0035] An initial interest point state determination model is determined, and the model is trained based on the second feature data and the estimated time state sequence to obtain the interest point state determination model.

[0036] In conjunction with the second aspect and the first implementation of the second aspect, in the second implementation of the present disclosure, the part that estimates the time state sequence of sample interest points based on the expired state time and the second feature data is configured as follows:

[0037] Calculate the cumulative feature data distribution of the sample interest points within the second preset historical time period;

[0038] The time point corresponding to when the accumulated feature data is the first set value is obtained;

[0039] Set the probability that the sample point of interest is in an expired state within the time period before the time point mentioned above to a first preset value.

[0040] Set the probability that the sample point of interest corresponding to the expiration time is in an expired state to a second preset value;

[0041] For the third preset historical time period between the time point corresponding to the first preset value of the accumulated feature data and the time of the expired state, interpolation is performed using the first preset value and the second preset value as endpoint probability values ​​to obtain the probability that the sample point of interest is in an expired state within the third preset historical time period. This results in the probability that the sample point of interest is in an expired state within the third preset historical time period showing a monotonically increasing trend, thereby obtaining the probability sequence of the state probability value of the sample point of interest as a function of time, i.e., the estimated time state sequence of the sample point of interest.

[0042] Combining the second aspect, the first implementation of the second aspect, and the second implementation of the second aspect, in the third implementation of the second aspect of this disclosure, the part of training the initial interest point state determination model based on the second feature data and the estimated time state sequence is configured as follows:

[0043] Obtain a training data set, wherein the training data set includes second feature training data obtained by sliding segmentation of the second feature data and the estimated time state sequence, and the estimated time state training sequence corresponding to the second feature training data;

[0044] Feature extraction is performed on the second feature training data to obtain the third feature training data corresponding to the second feature training data.

[0045] Using the second feature training data and the third feature training data as input, and the corresponding estimated time state training sequence as output, the initial interest point state determination model is trained to obtain the interest point state determination model.

[0046] Thirdly, embodiments of this disclosure provide an electronic device, including a memory and a processor. The memory stores one or more computer instructions that support a point-of-interest (POI) state determination device in executing the aforementioned POI state determination method. The processor is configured to execute the computer instructions stored in the memory. The POI state determination device may further include a communication interface for communicating with other devices or communication networks.

[0047] Fourthly, embodiments of this disclosure provide a computer-readable storage medium for storing computer instructions used by an interest point state determination device, including computer instructions for executing the aforementioned interest point state determination method for the interest point state determination device.

[0048] The technical solutions provided in this disclosure may have the following beneficial effects:

[0049] The aforementioned technical solution predicts the state probability of a target point of interest (POI) based on its feature data, and promptly determines the POI's state based on the relationship between its features and the probability of its state. This solution can quickly predict the POI's state, enabling timely notification to users, thereby effectively improving the service quality of internet platforms, ensuring a better user experience, and preventing users from incurring losses in time and travel distance.

[0050] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit the embodiments of this disclosure. Attached Figure Description

[0051] Other features, objects, and advantages of embodiments of this disclosure will become more apparent from the following detailed description of non-limiting implementations, taken in conjunction with the accompanying drawings. In the drawings:

[0052] Figure 1 A flowchart is shown illustrating a method for determining the state of a point of interest according to an embodiment of this disclosure;

[0053] Figure 2 A flowchart illustrating the overall process of a method for determining the state of points of interest according to an embodiment of the present disclosure is shown.

[0054] Figure 3A structural block diagram of a point of interest state determination apparatus according to an embodiment of the present disclosure is shown.

[0055] Figure 4 A structural block diagram of an electronic device according to an embodiment of the present disclosure is shown;

[0056] Figure 5 This is a schematic diagram of the structure of a computer system suitable for implementing the method for determining the state of points of interest according to an embodiment of the present disclosure. Detailed Implementation

[0057] Hereinafter, exemplary embodiments of the present disclosure will be described in detail with reference to the accompanying drawings to enable those skilled in the art to readily implement them. Furthermore, for clarity, portions unrelated to the description of the exemplary embodiments have been omitted from the drawings.

[0058] In embodiments disclosed herein, it should be understood that terms such as “comprising” or “having” are intended to indicate the presence of features, numbers, steps, behaviors, components, portions or combinations thereof disclosed herein, and are not intended to exclude the possibility that one or more other features, numbers, steps, behaviors, components, portions or combinations thereof are present or added.

[0059] It should also be noted that, unless otherwise specified, the embodiments and features described in this disclosure can be combined with each other. The embodiments of this disclosure will now be described in detail with reference to the accompanying drawings and examples.

[0060] The technical solution provided in this disclosure predicts the state probability of a target point of interest (POI) based on its feature data, thereby determining the POI's state in a timely manner based on the relationship between its features and its state. This solution can quickly predict the POI's state, enabling timely notification to users and effectively improving the service quality of internet platforms, ensuring a better user experience, and preventing users from incurring losses in time and distance.

[0061] Figure 1 A flowchart illustrating a method for determining the state of a point of interest according to an embodiment of this disclosure is shown, such as... Figure 1 As shown, the method for determining the state of a point of interest includes the following steps S101-S103:

[0062] In step S101, the first feature data of the target point of interest within a first preset historical time period is obtained;

[0063] In step S102, the first feature data is input into the pre-trained interest point state determination model to predict the state probability distribution of the target interest point within the first preset historical time period.

[0064] In step S103, the state probability corresponding to the preset time point in the state probability distribution within the first preset historical time period is compared with a preset probability threshold to determine the state of the target point of interest at the preset time point.

[0065] As mentioned above, with the development of the internet and data technology, more and more users are searching for and querying locations through internet platforms. However, in practical applications, the rate of change in the status of a location is relatively high, and it is difficult to notify every user of these changes in a timely manner. For example, if a restaurant closes, relocates, demolishes, or changes its name, even if its members are notified, users who are not members but want to dine there may not be informed in a timely manner. This causes great inconvenience to users' searches and may even result in time and travel losses for those who want to visit the restaurant. To solve the problem of not being able to notify users of location status changes in a timely manner, existing technologies mine location status changes. However, the mining speed of existing technologies is low, making it difficult to improve the service quality of internet platforms and guarantee the user experience.

[0066] To address the aforementioned issues, this implementation proposes a method for determining the state of a point of interest (POI). This method predicts the state probability of a POI based on its feature data, thereby determining the POI's state promptly based on the relationship between its features and its state. This technical solution can quickly predict the state of a POI, enabling timely notification to the user. This effectively improves the service quality of the internet platform, ensures a positive user experience, and avoids losses in terms of time and travel distance for users.

[0067] In one embodiment of this disclosure, the method for determining the state of points of interest can be applied to terminal computers, computing devices, electronic devices, servers, service clusters, etc., that can perform point of interest state determination processing.

[0068] In one embodiment of this disclosure, the point of interest refers to a geographical entity on a map that can be abstracted as a point of interest (POI), which may be queried by a user and used as a navigation destination, such as a hospital, park, shop, restaurant, etc.

[0069] In one embodiment of this disclosure, the target point of interest refers to a point of interest whose state needs to be determined.

[0070] In one embodiment of this disclosure, the first preset historical time period refers to a time period prior to the current time. Those skilled in the art can set the specific value of the first preset historical time period according to the needs of actual application. This disclosure does not impose any special limitations on it. For example, the first preset historical time period can be one year, two years, etc.

[0071] In one embodiment of this disclosure, the first feature data of the target point of interest refers to feature data used to characterize the state of the target point of interest, such as behavioral data of objects related to the target point of interest. For example, if the target point of interest is a restaurant, the first feature data of the target point of interest could be a sequence of user payment time data for that restaurant, etc.

[0072] In one embodiment of this disclosure, the point of interest state determination model refers to a pre-trained model used to determine the state of a point of interest. The input of the point of interest state determination model is the first feature data of the target point of interest within a first preset historical time period, and the output is the state probability distribution of the target point of interest within the first preset historical time period, that is, the probability value of the state of the target point of interest at each time point within the first preset historical time period. Here, the probability can be set to a second preset value, such as 1, when the state of the target point of interest is expired, and to a first preset value, such as 0, when the state of the target point of interest is not expired.

[0073] In one embodiment of this disclosure, the status of the point of interest can be divided into two states: expired state and non-expired state. The expired state refers to the point of interest being in an abnormal expired state, such as being closed, shut down, relocated, demolished, or renamed. The non-expired state refers to the point of interest being in a normal non-expired state, such as being in normal business operations.

[0074] In the above implementation, firstly, the first feature data of the target interest point whose state needs to be determined is obtained within a first preset historical time period. The first feature data is then input into a pre-trained interest point state determination model to predict the state probability distribution of the target interest point within the first preset historical time period. For a preset time point within the first preset historical time period, the state probability corresponding to that preset time point in the state probability distribution can be compared with a preset probability threshold to determine the state of the target interest point at that preset time point. For example, if the state probability corresponding to that preset time point in the state probability distribution is greater than the preset probability threshold, the state of the target interest point at that preset time point can be determined to be expired; if the state probability corresponding to that preset time point in the state probability distribution is less than or equal to the preset probability threshold, the state of the target interest point at that preset time point can be determined to be unexpired. The value of the preset probability threshold can be set according to the needs of the actual application. For example, if the state probability of the target interest point is in the range of (0, 1), the preset probability threshold can be set to 0.5, 0.6, 0.7, etc.

[0075] In one embodiment of this disclosure, the method may further include the following steps:

[0076] Train the interest point state determination model.

[0077] In one embodiment of this disclosure, the step of training the interest point state determination model may include the following steps:

[0078] The second feature data of the sample interest point is obtained, which is determined to be the expiration state time of the expiration state and the sample interest point is within a second preset historical time period before the expiration state time.

[0079] Based on the expired state time and the second feature data, the time state sequence of the sample interest points is estimated, and the estimated time state sequence of the sample interest points is obtained.

[0080] An initial interest point state determination model is determined, and the model is trained based on the second feature data and the estimated time state sequence to obtain the interest point state determination model.

[0081] In this embodiment, when training the point of interest state determination model, sample points of interest are first determined, and the expiration time for which the state of the sample point of interest is determined to be the expired state is obtained, as well as the second feature data of the sample point of interest within a second preset historical time period before the expiration time. Here, the sample point of interest refers to an interest point whose state has been determined and can be used as a sample; the expiration time refers to the time when the state of the sample point of interest can be determined to be expired based on user feedback or other determination methods; the second preset historical time period refers to a time period before the expiration time. Those skilled in the art can set the specific value of the second preset historical time period according to the needs of actual applications, and this disclosure does not impose any particular limitation on it. Similar to the first preset historical time period, the second preset historical time period can also be, for example, one year, two years, etc.; similar to the first feature data, the second feature data is also used to characterize the state of the sample point of interest, such as behavioral data of objects related to the sample point of interest, etc. For example, if the sample point of interest is a restaurant, the second feature data of the sample point of interest can be a sequence of user payment time data for that restaurant, etc. Then, based on the expired state time and the second feature data, the time state sequence of the sample interest point is roughly estimated, that is, the estimated time state sequence of the sample interest point. Here, the time state sequence refers to the probability sequence of the state probability value of the sample interest point as a function of time, that is, the probability that the sample interest point is in an expired state at different time points. The estimated time state sequence can be used as label data when training the interest point state determination model to achieve the effect of supervised learning, which is beneficial to improving the accuracy of interest point state change mining. Then, an initial interest point state determination model is determined. Based on the second feature data and the estimated time state sequence, the initial interest point state determination model is trained. After training, an interest point state determination model that can be used to determine the state of a certain interest point can be obtained. Here, the initial interest point state determination model can be, for example, a recurrent RNN neural network model, a gated recurrent unit (GRU) network model, etc.

[0082] In one embodiment of this disclosure, the step of estimating the time state sequence of sample points of interest based on the expired state time and the second feature data may include the following steps:

[0083] Calculate the cumulative feature data distribution of the sample interest points within the second preset historical time period;

[0084] The time point corresponding to when the accumulated feature data is the first set value is obtained;

[0085] Set the probability that the sample point of interest is in an expired state within the time period before the time point mentioned above to a first preset value.

[0086] Set the probability that the sample point of interest corresponding to the expiration time is in an expired state to a second preset value;

[0087] For the third preset historical time period between the time point corresponding to the first preset value of the accumulated feature data and the time of the expired state, interpolation is performed using the first preset value and the second preset value as endpoint probability values ​​to obtain the probability that the sample point of interest is in an expired state within the third preset historical time period. This results in the probability that the sample point of interest is in an expired state within the third preset historical time period showing a monotonically increasing trend, thereby obtaining the probability sequence of the state probability value of the sample point of interest as a function of time, i.e., the estimated time state sequence of the sample point of interest.

[0088] In this embodiment, when estimating the time state sequence of sample points of interest, the cumulative feature data distribution of the sample points of interest within the second preset historical time period is first calculated. For example, the cumulative feature data of the sample points of interest at each time point within the second preset historical time period can be calculated based on the Poisson cumulative distribution function. Considering that when the cumulative feature data is not a first set value, the sample points of interest should be in a normal or unexpired state, the probability value of the sample points of interest being in an expired state within the time period before the time point corresponding to the first set value of the cumulative feature data in the second preset historical time period can be set as the first set value. Considering that the state of the sample points of interest has been determined to be expired at the expired state time, the expired state time can be... The probability value of the corresponding sample point of interest being in an expired state is set to a second preset value. For the remaining time period, that is, the third preset historical time period between the time point corresponding to the cumulative feature data being at the first preset value and the time of the expired state, interpolation can be performed based on the probability values ​​of the first and second preset values, and the time decay factor is considered to obtain the probability of the sample point of interest being in an expired state within the third preset historical time period. This makes the probability of the sample point of interest being in an expired state within the third preset historical time period show a monotonically increasing trend. In this way, based on the probability of the sample point of interest being in an expired state within the third preset historical time period, the probability sequence of the state probability value of the sample point of interest as a function of time can be obtained, that is, the estimated time state sequence of the sample point of interest.

[0089] In one embodiment of this disclosure, the step of training the initial interest point state determination model based on the second feature data and the estimated time state sequence may include the following steps:

[0090] Obtain a training data set, wherein the training data set includes second feature training data obtained by sliding segmentation of the second feature data and the estimated time state sequence, and the estimated time state training sequence corresponding to the second feature training data;

[0091] Feature extraction is performed on the second feature training data to obtain the third feature training data corresponding to the second feature training data.

[0092] Using the second feature training data and the third feature training data as input, and the corresponding estimated time state training sequence as output, the initial interest point state determination model is trained to obtain the interest point state determination model.

[0093] In this embodiment, when training the initial interest point state determination model based on the second feature data and the estimated time state sequence, a training data set can be determined first. This training data set includes second feature training data obtained by sliding segmentation of the second feature data and the estimated time state sequence, and the estimated time state training sequence corresponding to the second feature training data. The sliding segmentation refers to data segmentation using a sliding time window with a specific length and a specific sliding step size. For example, if the second preset historical time period is one year, then both the second feature data and the estimated time state sequence are sequences with a dimension of 365*1. Assuming the length of the sliding time window is one month (60 days) and the sliding step size is one day, then after sliding segmentation of the second feature data and the estimated time state sequence using the sliding time window, 306 pairs of sequences with a dimension of 60*1 can be obtained, i.e., 306 second feature training data with a dimension of 60*1 and 306 estimated time state training sequences with a dimension of 60*1 corresponding to the second feature training data. Then, feature extraction is performed on the second feature training data to obtain the third feature training data corresponding to the second feature training data. Finally, both the second and third feature training data are used as inputs, and the corresponding estimated time state training sequence is used as output to train the initial interest point state determination model. After training, the interest point state determination model is obtained, thus establishing the relationship between the target interest point features and the probability of the target interest point state. It learns the feature data change trend that is as close as possible to the actual state change time of the target interest point, so that the state change of the target interest point can be captured in a timely manner. A loss function can be set to measure the difference between the output of the interest point state determination model and the estimated time state training sequence used as label data. The model parameters of the interest point state determination model are adjusted according to the value of the loss function. When the loss function reaches its minimum, the interest point state determination model is considered to have completed training and reached convergence.

[0094] In one embodiment of this disclosure, the step of extracting features from the second feature training data to obtain third feature training data corresponding to the second feature training data may include the following steps:

[0095] Extract local features from the training data for the second feature;

[0096] Extract the global features from the training data for the second feature;

[0097] The local and global features of the second feature training data are combined to form the third feature training data corresponding to the second feature training data.

[0098] In this embodiment, when extracting features from the second feature training data, local features are first extracted to obtain the trend of change of the second feature training data on a small time scale, such as local increase or decrease, local stability or fluctuation, etc.; then, global features are extracted to obtain the trend of change of the second feature training data on an overall time scale, such as overall increase or decrease, overall stability or fluctuation, etc.; finally, the local and global features of the second feature training data are combined, and the combined features are combined to obtain the third feature training data corresponding to the second feature training data, thereby enriching the trend information of feature data, realizing multi-granular feature extraction of the interest point state determination model, reducing noise interference, and further improving the accuracy of interest point state change mining. If the dimension of the second feature training data is 60*1, then the dimensions of the local and global features of the second feature training data are both 60*1, while the dimension of the third feature training data is 60*2. The extraction of local and global features can be achieved using a gated recurrent unit network model.

[0099] In one embodiment of this disclosure, the method may further include the following steps:

[0100] When the target point of interest is in an expired state, a preset operation is performed.

[0101] In this implementation, after obtaining the status of the target point of interest, corresponding preset operations can be performed based on the status of the target point of interest. For example, when it is determined that the status of the target point of interest is abnormal or expired, in order to avoid loss of time and travel for users, effectively improve the service quality of the Internet platform, and ensure the user experience, the status of the target point of interest can be notified in map applications, lifestyle applications, transportation applications, or through other channels, etc.

[0102] Figure 2 This diagram illustrates the overall flowchart of a method for determining the state of points of interest according to an embodiment of the present disclosure, as follows: Figure 2As shown, firstly, using the second feature data of the sample interest points within a second preset historical time period before the expiration time determined to be the expiration state, the time state sequence of the sample interest points is estimated, resulting in an estimated time state sequence. Then, local and global features of the second feature data are extracted respectively, and combined to obtain a combined feature. Combining the local, global, and combined features yields the third feature data. Then, the second feature data, the third feature data, and the estimated time state sequence are used as training data to train the interest point state determination model. During training, a loss function can be set to measure the difference between the output of the interest point state determination model, i.e., the predicted time state sequence, and the estimated time state training sequence used as training label data, in order to adjust the model parameters of the interest point state determination model so that the loss function is minimized and the interest point state determination model reaches convergence. In application, the first feature data of the target interest point within a first preset historical time period can be obtained first. The first feature data is then input into the trained interest point state determination model to predict the state probability distribution of the target interest point within the first preset historical time period. For a certain time point within the first preset historical time period, if the state probability corresponding to it in the state probability distribution is greater than a preset probability threshold, it can be determined that the target interest point is in an expired state at this time. If it is less than or equal to the preset probability threshold, it can be determined that the target interest point is in an unexpired state at this time.

[0103] The following are embodiments of the apparatus disclosed herein, which can be used to execute embodiments of the method disclosed herein.

[0104] Figure 3 This diagram illustrates a structural block diagram of a point-of-interest (POI) state determination apparatus according to an embodiment of the present disclosure. This apparatus can be implemented as part or all of an electronic device through software, hardware, or a combination of both. Figure 3 As shown, the point of interest state determination device includes:

[0105] The acquisition module 301 is configured to acquire the first feature data of the target point of interest within a first preset historical time period;

[0106] Prediction module 302 is configured to input the first feature data into a pre-trained interest point state determination model to predict the state probability distribution of the target interest point within the first preset historical time period.

[0107] The determination module 303 is configured to compare the state probability corresponding to the preset time point in the state probability distribution within the first preset historical time period with a preset probability threshold to determine the state of the target point of interest at the preset time point.

[0108] As mentioned above, with the development of the internet and data technology, more and more users are searching for and querying locations through internet platforms. However, in practical applications, the rate of change in the status of a location is relatively high, and it is difficult to notify every user of these changes in a timely manner. For example, if a restaurant closes, relocates, demolishes, or changes its name, even if its members are notified, users who are not members but want to dine there may not be informed in a timely manner. This causes great inconvenience to users' searches and may even result in time and travel losses for those who want to visit the restaurant. To solve the problem of not being able to notify users of location status changes in a timely manner, existing technologies mine location status changes. However, the mining speed of existing technologies is low, making it difficult to improve the service quality of internet platforms and guarantee the user experience.

[0109] In view of the above problems, this embodiment proposes a point-of-interest (POI) state determination device. This device predicts the state probability of a POI based on its feature data, and determines the POI's state promptly based on the relationship between its features and its state. This technical solution can quickly predict the state of a POI, thereby enabling timely notification to the user. This effectively improves the service quality of the internet platform, ensures a good user experience, and avoids losses in terms of time and travel distance for users.

[0110] In one embodiment of this disclosure, the point of interest state determination device can be implemented as a terminal computer, computing device, electronic device, server, service cluster, etc., capable of performing point of interest state determination processing.

[0111] In one embodiment of this disclosure, the point of interest refers to a geographical entity on a map that can be abstracted as a point of interest (POI), which may be queried by a user and used as a navigation destination, such as a hospital, park, shop, restaurant, etc.

[0112] In one embodiment of this disclosure, the target point of interest refers to a point of interest whose state needs to be determined.

[0113] In one embodiment of this disclosure, the first preset historical time period refers to a time period prior to the current time. Those skilled in the art can set the specific value of the first preset historical time period according to the needs of actual application. This disclosure does not impose any special limitations on it. For example, the first preset historical time period can be one year, two years, etc.

[0114] In one embodiment of this disclosure, the first feature data of the target point of interest refers to feature data used to characterize the state of the target point of interest, such as behavioral data of objects related to the target point of interest. For example, if the target point of interest is a restaurant, the first feature data of the target point of interest could be a sequence of user payment time data for that restaurant, etc.

[0115] In one embodiment of this disclosure, the point of interest state determination model refers to a pre-trained model used to determine the state of a point of interest. The input of the point of interest state determination model is the first feature data of the target point of interest within a first preset historical time period, and the output is the state probability distribution of the target point of interest within the first preset historical time period, that is, the probability value of the state of the target point of interest at each time point within the first preset historical time period. Here, the probability can be set to a second preset value, such as 1, when the state of the target point of interest is expired, and to a first preset value, such as 0, when the state of the target point of interest is not expired.

[0116] In one embodiment of this disclosure, the status of the point of interest can be divided into two states: expired state and non-expired state. The expired state refers to the point of interest being in an abnormal expired state, such as being closed, shut down, relocated, demolished, or renamed. The non-expired state refers to the point of interest being in a normal non-expired state, such as being in normal business operations.

[0117] In the above implementation, firstly, the first feature data of the target interest point whose state needs to be determined is obtained within a first preset historical time period. The first feature data is then input into a pre-trained interest point state determination model to predict the state probability distribution of the target interest point within the first preset historical time period. For a preset time point within the first preset historical time period, the state probability corresponding to that preset time point in the state probability distribution can be compared with a preset probability threshold to determine the state of the target interest point at that preset time point. For example, if the state probability corresponding to that preset time point in the state probability distribution is greater than the preset probability threshold, the state of the target interest point at that preset time point can be determined to be expired; if the state probability corresponding to that preset time point in the state probability distribution is less than or equal to the preset probability threshold, the state of the target interest point at that preset time point can be determined to be unexpired. The value of the preset probability threshold can be set according to the needs of the actual application. For example, if the state probability of the target interest point is in the range of (0, 1), the preset probability threshold can be set to 0.5, 0.6, 0.7, etc.

[0118] In one embodiment of this disclosure, the apparatus may further include:

[0119] The training module is configured to train the interest point state determination model.

[0120] In one embodiment of this disclosure, the training module is configured as follows:

[0121] The second feature data of the sample interest point is obtained, which is determined to be the expiration state time of the expiration state and the sample interest point is within a second preset historical time period before the expiration state time.

[0122] Based on the expired state time and the second feature data, the time state sequence of the sample interest points is estimated, and the estimated time state sequence of the sample interest points is obtained.

[0123] An initial interest point state determination model is determined, and the model is trained based on the second feature data and the estimated time state sequence to obtain the interest point state determination model.

[0124] In this embodiment, when training the point of interest state determination model, sample points of interest are first determined, and the expiration time for which the state of the sample point of interest is determined to be the expired state is obtained, as well as the second feature data of the sample point of interest within a second preset historical time period before the expiration time. Here, the sample point of interest refers to an interest point whose state has been determined and can be used as a sample; the expiration time refers to the time when the state of the sample point of interest can be determined to be expired based on user feedback or other determination methods; the second preset historical time period refers to a time period before the expiration time. Those skilled in the art can set the specific value of the second preset historical time period according to the needs of actual applications, and this disclosure does not impose any particular limitation on it. Similar to the first preset historical time period, the second preset historical time period can also be, for example, one year, two years, etc.; similar to the first feature data, the second feature data is also used to characterize the state of the sample point of interest, such as behavioral data of objects related to the sample point of interest, etc. For example, if the sample point of interest is a restaurant, the second feature data of the sample point of interest can be a sequence of user payment time data for that restaurant, etc. Then, based on the expired state time and the second feature data, the time state sequence of the sample interest point is roughly estimated, that is, the estimated time state sequence of the sample interest point. Here, the time state sequence refers to the probability sequence of the state probability value of the sample interest point as a function of time, that is, the probability that the sample interest point is in an expired state at different time points. The estimated time state sequence can be used as label data when training the interest point state determination model to achieve the effect of supervised learning, which is beneficial to improving the accuracy of interest point state change mining. Then, an initial interest point state determination model is determined. Based on the second feature data and the estimated time state sequence, the initial interest point state determination model is trained. After training, an interest point state determination model that can be used to determine the state of a certain interest point can be obtained. Here, the initial interest point state determination model can be, for example, a recurrent RNN neural network model, a gated recurrent unit (GRU) network model, etc.

[0125] In one embodiment of this disclosure, the part that estimates the time state sequence of sample interest points based on the expired state time and the second feature data can be configured as follows:

[0126] Calculate the cumulative feature data distribution of the sample interest points within the second preset historical time period;

[0127] The time point corresponding to when the accumulated feature data is the first set value is obtained;

[0128] Set the probability that the sample point of interest is in an expired state within the time period before the time point mentioned above to a first preset value.

[0129] Set the probability that the sample point of interest corresponding to the expiration time is in an expired state to a second preset value;

[0130] For the third preset historical time period between the time point corresponding to the first preset value of the accumulated feature data and the time of the expired state, interpolation is performed using the first preset value and the second preset value as endpoint probability values ​​to obtain the probability that the sample point of interest is in an expired state within the third preset historical time period. This results in the probability that the sample point of interest is in an expired state within the third preset historical time period showing a monotonically increasing trend, thereby obtaining the probability sequence of the state probability value of the sample point of interest as a function of time, i.e., the estimated time state sequence of the sample point of interest.

[0131] In this embodiment, when estimating the time state sequence of sample points of interest, the cumulative feature data distribution of the sample points of interest within the second preset historical time period is first calculated. For example, the cumulative feature data of the sample points of interest at each time point within the second preset historical time period can be calculated based on the Poisson cumulative distribution function. Considering that when the cumulative feature data is not a first set value, the sample points of interest should be in a normal or unexpired state, the probability value of the sample points of interest being in an expired state within the time period before the time point corresponding to the first set value of the cumulative feature data in the second preset historical time period can be set as the first set value. Considering that the state of the sample points of interest has been determined to be expired at the expired state time, the expired state time can be... The probability value of the corresponding sample point of interest being in an expired state is set to a second preset value. For the remaining time period, that is, the third preset historical time period between the time point corresponding to the cumulative feature data being at the first preset value and the time of the expired state, interpolation can be performed based on the probability values ​​of the first and second preset values, and the time decay factor is considered to obtain the probability of the sample point of interest being in an expired state within the third preset historical time period. This makes the probability of the sample point of interest being in an expired state within the third preset historical time period show a monotonically increasing trend. In this way, based on the probability of the sample point of interest being in an expired state within the third preset historical time period, the probability sequence of the state probability value of the sample point of interest as a function of time can be obtained, that is, the estimated time state sequence of the sample point of interest.

[0132] In one embodiment of this disclosure, the part that trains the initial interest point state determination model based on the second feature data and the estimated time state sequence can be configured as follows:

[0133] Obtain a training data set, wherein the training data set includes second feature training data obtained by sliding segmentation of the second feature data and the estimated time state sequence, and the estimated time state training sequence corresponding to the second feature training data;

[0134] Feature extraction is performed on the second feature training data to obtain the third feature training data corresponding to the second feature training data.

[0135] Using the second feature training data and the third feature training data as input, and the corresponding estimated time state training sequence as output, the initial interest point state determination model is trained to obtain the interest point state determination model.

[0136] In this embodiment, when training the initial interest point state determination model based on the second feature data and the estimated time state sequence, a training data set can be determined first. This training data set includes second feature training data obtained by sliding segmentation of the second feature data and the estimated time state sequence, and the estimated time state training sequence corresponding to the second feature training data. The sliding segmentation refers to data segmentation using a sliding time window with a specific length and a specific sliding step size. For example, if the second preset historical time period is one year, then both the second feature data and the estimated time state sequence are sequences with a dimension of 365*1. Assuming the length of the sliding time window is one month (60 days) and the sliding step size is one day, then after sliding segmentation of the second feature data and the estimated time state sequence using the sliding time window, 306 pairs of sequences with a dimension of 60*1 can be obtained, i.e., 306 second feature training data with a dimension of 60*1 and 306 estimated time state training sequences with a dimension of 60*1 corresponding to the second feature training data. Then, feature extraction is performed on the second feature training data to obtain the third feature training data corresponding to the second feature training data. Finally, both the second and third feature training data are used as inputs, and the corresponding estimated time state training sequence is used as output to train the initial interest point state determination model. After training, the interest point state determination model is obtained, thus establishing the relationship between the target interest point features and the probability of the target interest point state. It learns the feature data change trend that is as close as possible to the actual state change time of the target interest point, so that the state change of the target interest point can be captured in a timely manner. A loss function can be set to measure the difference between the output of the interest point state determination model and the estimated time state training sequence used as label data. The model parameters of the interest point state determination model are adjusted according to the value of the loss function. When the loss function reaches its minimum, the interest point state determination model is considered to have completed training and reached convergence.

[0137] In one embodiment of this disclosure, the step of performing secondary feature extraction on the second feature training data to obtain the third feature training data corresponding to the second feature training data can be configured as follows:

[0138] Extract local features from the training data for the second feature;

[0139] Extract the global features from the second feature training data;

[0140] The local and global features of the second feature training data are combined to form the third feature training data corresponding to the second feature training data.

[0141] In this embodiment, when extracting features from the second feature training data, local features are first extracted to obtain the trend of change of the second feature training data on a small time scale, such as local increase or decrease, local stability or fluctuation, etc.; then, global features are extracted to obtain the trend of change of the second feature training data on an overall time scale, such as overall increase or decrease, overall stability or fluctuation, etc.; finally, the local and global features of the second feature training data are combined, and the combined features are combined to obtain the third feature training data corresponding to the second feature training data, thereby enriching the trend information of feature data, realizing multi-granular feature extraction of the interest point state determination model, reducing noise interference, and further improving the accuracy of interest point state change mining. If the dimension of the second feature training data is 60*1, then the dimensions of the local and global features of the second feature training data are both 60*1, while the dimension of the third feature training data is 60*2. The extraction of local and global features can be achieved using a gated recurrent unit network model.

[0142] In one embodiment of this disclosure, the apparatus may further include:

[0143] The execution module is configured to perform a preset operation when the target point of interest is in an expired state.

[0144] In this implementation, after obtaining the status of the target point of interest, corresponding preset operations can be performed based on the status of the target point of interest. For example, when it is determined that the status of the target point of interest is abnormal or expired, in order to avoid loss of time and travel for users, effectively improve the service quality of the Internet platform, and ensure the user experience, the status of the target point of interest can be notified in map applications, lifestyle applications, transportation applications, or through other channels, etc.

[0145] This disclosure also discloses an electronic device, Figure 4 This diagram illustrates a structural block diagram of an electronic device according to an embodiment of the present disclosure, such as... Figure 4 As shown, the electronic device 400 includes a memory 401 and a processor 402; wherein,

[0146] The memory 401 is used to store one or more computer instructions, wherein the one or more computer instructions are executed by the processor 402 to implement the above method steps.

[0147] Figure 5 This is a schematic diagram of the structure of a computer system suitable for implementing the method for determining the state of points of interest according to an embodiment of the present disclosure.

[0148] like Figure 5 As shown, the computer system 500 includes a processing unit 501, which can execute various processes described above based on a program stored in a read-only memory (ROM) 502 or a program loaded from a storage section 508 into a random access memory (RAM) 503. The RAM 503 also stores various programs and data required for the operation of the system 500. The processing unit 501, ROM 502, and RAM 503 are interconnected via a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0149] The following components are connected to I / O interface 505: an input section 506 including a keyboard, mouse, etc.; an output section 507 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 508 including a hard disk, etc.; and a communication section 509 including a network interface card such as a LAN card, modem, etc. The communication section 509 performs communication processing via a network such as the Internet. A drive 510 is also connected to I / O interface 505 as needed. A removable medium 511, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on drive 510 as needed so that computer programs read from it can be installed into storage section 508 as needed. The processing unit 501 can be implemented as a CPU, GPU, TPU, FPGA, NPU, etc.

[0150] In particular, according to embodiments of this disclosure, the methods described above can be implemented as computer software programs. For example, embodiments of this disclosure include a computer program product comprising a computer program tangibly embodied on a readable medium thereof, the computer program containing program code for performing the accessibility information determination method. In such embodiments, the computer program can be downloaded and installed from a network via communication section 509, and / or installed from removable medium 511.

[0151] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0152] The units or modules described in the embodiments of this disclosure can be implemented in software or hardware. The described units or modules can also be located in a processor, and the names of these units or modules do not necessarily constitute a limitation on the unit or module itself.

[0153] In another aspect, embodiments of this disclosure also provide a computer-readable storage medium, which may be a computer-readable storage medium included in the apparatus described in the above embodiments; or it may be a standalone computer-readable storage medium not assembled into a device. The computer-readable storage medium stores one or more programs, which are used by one or more processors to perform the methods described in embodiments of this disclosure.

[0154] The above description is merely a preferred embodiment of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in the embodiments of this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the inventive concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in the embodiments of this disclosure.

Claims

1. A method for determining the state of a point of interest, wherein, include: Acquire the first feature data of the target point of interest within a first preset historical time period; The first feature data is input into a pre-trained interest point state determination model to predict the state probability distribution of the target interest point within the first preset historical time period. The pre-trained interest point state determination model is trained based on the second feature data and the estimated time state sequence. The second feature data consists of feature data of the sample interest point within a second preset historical time period before its expiration time. The estimated time state sequence is obtained by estimating the time state sequence of the sample interest point based on the expiration time when the sample interest point is determined to be in an expired state and the second feature data. The state probability corresponding to the preset time point in the state probability distribution within the first preset historical time period is compared with a preset probability threshold to determine the state of the target point of interest at the preset time point.

2. The method according to claim 1, wherein, The method further includes: Acquire the expiration time of the sample interest point when it is determined to be in an expired state, and the second feature data of the sample interest point within a second preset historical time period before the expiration time; Based on the expired state time and the second feature data, the time state sequence of the sample interest points is estimated, and the estimated time state sequence of the sample interest points is obtained. An initial interest point state determination model is determined, and the model is trained based on the second feature data and the estimated time state sequence to obtain the interest point state determination model.

3. The method according to claim 2, wherein, The step of estimating the temporal state sequence of sample interest points based on the expired state time and the second feature data includes: Calculate the cumulative feature data distribution of the sample interest points within the second preset historical time period; The time point corresponding to when the accumulated feature data is the first set value is obtained; Set the probability that the sample point of interest is in an expired state within the time period before the time point mentioned above to a first preset value. Set the probability that the sample point of interest corresponding to the expiration time is in an expired state to a second preset value; For the third preset historical time period between the time point corresponding to the first preset value of the accumulated feature data and the time of the expired state, interpolation is performed using the first preset value and the second preset value as endpoint probability values ​​to obtain the probability that the sample point of interest is in an expired state within the third preset historical time period. This results in the probability that the sample point of interest is in an expired state within the third preset historical time period showing a monotonically increasing trend, thereby obtaining the probability sequence of the state probability value of the sample point of interest as a function of time, i.e., the estimated time state sequence of the sample point of interest.

4. The method according to claim 2 or 3, wherein, The step of training the initial interest point state determination model based on the second feature data and the estimated time state sequence includes: Obtain a training data set, wherein the training data set includes second feature training data obtained by sliding segmentation of the second feature data and the estimated time state sequence, and the estimated time state training sequence corresponding to the second feature training data; Feature extraction is performed on the second feature training data to obtain the third feature training data corresponding to the second feature training data. Using the second feature training data and the third feature training data as input, and the corresponding estimated time state training sequence as output, the initial interest point state determination model is trained to obtain the interest point state determination model.

5. The method according to claim 4, wherein, The step of extracting features from the second feature training data to obtain third feature training data corresponding to the second feature training data includes: Extract local features from the training data for the second feature; Extract the global features from the training data for the second feature; The local and global features of the second feature training data are combined to form the third feature training data corresponding to the second feature training data.

6. A device for determining the state of a point of interest, comprising: The acquisition module is configured to acquire the first feature data of the target point of interest within a first preset historical time period; The prediction module is configured to input the first feature data into a pre-trained interest point state determination model to predict the state probability distribution of the target interest point within a first preset historical time period; wherein, the pre-trained interest point state determination model is obtained by training an initial interest point state determination model based on the second feature data and the estimated time state sequence; the second feature data is the feature data of the sample interest point within a second preset historical time period before the expiration time, and the estimated time state sequence is obtained by estimating the time state sequence of the sample interest point based on the expiration time when the sample interest point is determined to be in an expired state and the second feature data. The determination module is configured to compare the state probability corresponding to the preset time point in the state probability distribution within the first preset historical time period with a preset probability threshold to determine the state of the target point of interest at the preset time point.

7. The apparatus according to claim 6, further comprising a training module, wherein, The training module is configured as follows: Acquire the expiration time of the sample interest point when it is determined to be in an expired state, and the second feature data of the sample interest point within a second preset historical time period before the expiration time; Based on the expired state time and the second feature data, the time state sequence of the sample interest points is estimated, and the estimated time state sequence of the sample interest points is obtained. An initial interest point state determination model is determined, and the model is trained based on the second feature data and the estimated time state sequence to obtain the interest point state determination model.

8. The apparatus according to claim 7, wherein, The portion that estimates the time state sequence of sample interest points based on the expired state time and the second feature data is configured as follows: Calculate the cumulative feature data distribution of the sample interest points within the second preset historical time period; The time point corresponding to when the accumulated feature data is the first set value is obtained; Set the probability that the sample point of interest is in an expired state within the time period before the time point mentioned above to a first preset value. Set the probability that the sample point of interest corresponding to the expiration time is in an expired state to a second preset value; For the third preset historical time period between the time point corresponding to the first preset value of the accumulated feature data and the time of the expired state, interpolation is performed using the first preset value and the second preset value as endpoint probability values ​​to obtain the probability that the sample point of interest is in an expired state within the third preset historical time period. This results in the probability that the sample point of interest is in an expired state within the third preset historical time period showing a monotonically increasing trend, thereby obtaining the probability sequence of the state probability value of the sample point of interest as a function of time, i.e., the estimated time state sequence of the sample point of interest.

9. The apparatus according to claim 7 or 8, wherein, The part of training the initial interest point state determination model based on the second feature data and the estimated time state sequence is configured as follows: Obtain a training data set, wherein the training data set includes second feature training data obtained by sliding segmentation of the second feature data and the estimated time state sequence, and the estimated time state training sequence corresponding to the second feature training data; Feature extraction is performed on the second feature training data to obtain the third feature training data corresponding to the second feature training data. Using the second feature training data and the third feature training data as input, and the corresponding estimated time state training sequence as output, the initial interest point state determination model is trained to obtain the interest point state determination model.

10. An electronic device, characterized in that, The method includes a memory and at least one processor; wherein the memory is used to store one or more computer instructions, wherein the one or more computer instructions are executed by the at least one processor to implement the steps of the method according to any one of claims 1-5.

Images