Interest point state identification method and device

Abstract

The invention provides an interest point state recognition method and device, and relates to the technical field of map navigation, in particular to the technical field of navigation and position service, interest point state recognition and deep learning. According to one specific embodiment, the method comprises the steps that information of two adjacent navigation behaviors is obtained, and the two adjacent navigation behaviors comprise the navigation behavior of arriving at a first interest point and the navigation behavior of arriving at a second interest point from the first interest point; analyzing information of two adjacent navigation behaviors, and determining navigation behavior time sequence data; obtaining navigation behavior time sequence characteristics based on the navigation behavior time sequence data; the navigation behavior time sequence characteristics are input into a pre-trained state discrimination model, discrimination scores of the navigation behavior time sequence characteristics are obtained, and the state discrimination model is used for discriminating the states of the interest points; and determining the state of the first interest point based on the discrimination score of the navigation behavior time sequence characteristics.

Description

Technical Field

[0001] The present disclosure relates to the technical field of map navigation, and particularly to the technical fields of navigation and location services, point of interest (POI) status recognition, and deep learning. Background Art

[0002] POI (Point of Interest) data is a core element in map services, which is used to depict locations such as food restaurants, supermarkets, scenic spots, and institutions in the real world to meet the retrieval and navigation needs of users. However, the currency (i.e., timeliness and accuracy of data) of POI data is an important dimension for evaluating its quality. Therefore, it is necessary to perform POI status mining to update the POI status in a timely manner.

[0003] Currently, POI status mining is mainly carried out in the following three ways.

[0004] Firstly, image difference: Discover the change of status through the change of signs or OCR (Optical Character Recognition) in front and back images. The image recognizes the opening and closing data of stores, and combines the opening and closing characteristics to perform POI status mining.

[0005] Secondly, Wifi trend mining: Combine the temporal change of the number of users connected to Wifi to perform POI status mining.

[0006] Thirdly, user feedback: Perform POI status mining through the platform feedback information of users. Summary of the Invention

[0007] Embodiments of the present disclosure propose a method, device, equipment, storage medium, and program product for POI status recognition.

[0008] In a first aspect, embodiments of the present disclosure propose a method for POI status recognition, including: obtaining information on two adjacent navigation behaviors, where the two adjacent navigation behaviors include a navigation behavior to reach a first POI and a navigation behavior from the first POI to a second POI; analyzing the information on the two adjacent navigation behaviors to determine navigation behavior time series data; obtaining navigation behavior time series features based on the navigation behavior time series data; inputting the navigation behavior time series features into a pre-trained status discrimination model to obtain a discrimination score of the navigation behavior time series features, where the status discrimination model is used to discriminate the status of a POI; and determining the status of the first POI based on the discrimination score of the navigation behavior time series features.

[0009] Second aspect, an embodiment of the present disclosure provides a method for training a state discrimination model, including: obtaining training samples, where the training samples include information on two adjacent navigation behaviors of a sample user and a sample interest point state label, and the two adjacent navigation behaviors of the sample user include a navigation behavior of the sample user arriving at a sample first interest point and a navigation behavior of arriving at a sample second interest point from the sample first interest point, and the sample interest point state label is the state of the sample first interest point; analyzing the information on the two adjacent navigation behaviors of the sample user to determine sample navigation behavior time series data; obtaining sample navigation behavior time series features based on the sample navigation behavior time series data; inputting the sample navigation behavior time series features into a time series model to obtain a discrimination score of the sample navigation behavior time series features; calculating a loss based on the discrimination score of the sample navigation behavior time series features and the sample interest point state label; and adjusting the parameters of the time series model based on the loss to obtain a state discrimination model.

[0010] Third aspect, an embodiment of the present disclosure provides an interest point state recognition device, including: a first obtaining module configured to obtain information on two adjacent navigation behaviors, where the two adjacent navigation behaviors include a navigation behavior of arriving at a first interest point and a navigation behavior of arriving at a second interest point from the first interest point; an analysis module configured to analyze the information on the two adjacent navigation behaviors to determine navigation behavior time series data; a second obtaining module configured to obtain navigation behavior time series features based on the navigation behavior time series data; a discrimination module configured to input the navigation behavior time series features into a pre-trained state discrimination model to obtain a discrimination score of the navigation behavior time series features, where the state discrimination model is used to discriminate the state of an interest point; and a determination module configured to determine the state of the first interest point based on the discrimination score of the navigation behavior time series features.

[0011] Fourth aspect, an embodiment of the present disclosure provides a state discrimination model training device, including: a first obtaining module configured to obtain training samples, where the training samples include information on two adjacent navigation behaviors of a sample user and a sample interest point state label, and the two adjacent navigation behaviors of the sample user include a navigation behavior of the sample user arriving at a sample first interest point and a navigation behavior of arriving at a sample second interest point from the sample first interest point, and the sample interest point state label is the state of the sample first interest point; an analysis module configured to analyze the information on the two adjacent navigation behaviors of the sample user to determine sample navigation behavior time series data; a second obtaining module configured to obtain sample navigation behavior time series features based on the sample navigation behavior time series data; a discrimination module configured to input the sample navigation behavior time series features into a time series model to obtain a discrimination score of the sample navigation behavior time series features; a calculation module configured to calculate a loss based on the discrimination score of the sample navigation behavior time series features and the sample interest point state label; and an adjustment module configured to adjust the parameters of the time series model based on the loss to obtain a state discrimination model.

[0012] In a fifth aspect, an embodiment of the present disclosure provides an electronic device, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein, the memory stores instructions executable by the at least one processor, and when the instructions are executed by the at least one processor, the at least one processor is enabled to execute the method described in the first aspect or the second aspect.

[0013] In a sixth aspect, an embodiment of the present disclosure provides a non-transitory computer-readable storage medium storing computer instructions for causing a computer to execute the method described in the first aspect or the second aspect.

[0014] In a seventh aspect, an embodiment of the present disclosure provides a computer program product including a computer program, which when executed by a processor implements the method described in the first aspect or the second aspect.

[0015] The key or important features of the embodiments of the present disclosure are not used to limit the scope of the present disclosure. Other features of the present disclosure will become easily understandable through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Other features, objects, and advantages of the present disclosure will become more apparent by reading the detailed description of the non-limiting embodiments with reference to the following drawings. The drawings are used to better understand the solution and do not constitute a limitation to the present disclosure. Among them: Figure 1 is a flowchart of an embodiment of the method for training a state discrimination model according to the present disclosure; Figure 2 is a flowchart of another embodiment of the method for training a state discrimination model according to the present disclosure; Figure 3 is a flowchart of an embodiment of the method for identifying the state of a point of interest according to the present disclosure; Figure 4 is a flowchart of another embodiment of the method for identifying the state of a point of interest according to the present disclosure; Figure 5 is a schematic diagram of a navigation behavior acquisition module; Figure 6 is a schematic diagram of a feature extraction module; Figure 7 is a schematic diagram of a multi-temporal model mining module; Figure 8 is a schematic structural diagram of an embodiment of the state discrimination model training apparatus according to the present disclosure; Figure 9 is a schematic structural diagram of an embodiment of the point of interest state recognition apparatus according to the present disclosure; Figure 10It is a block diagram of an electronic device for implementing the point of interest status recognition method of the embodiments of the present disclosure. Detailed implementation manners

[0017] The following describes exemplary embodiments of the present disclosure with reference to the accompanying drawings. Various details of the embodiments of the present disclosure are included to facilitate understanding, and they should be considered merely exemplary. Therefore, those of ordinary skill in the art should recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0018] It should be noted that, without conflict, the embodiments in the present disclosure and the features in the embodiments can be combined with each other. The present disclosure will be described in detail below with reference to the accompanying drawings and in conjunction with the embodiments.

[0019] Figure 1 Flow 100 of an embodiment of the state discrimination model training method according to the present disclosure is shown. The state discrimination model training method includes the following steps: Step 101, obtain training samples.

[0020] In this embodiment, the execution subject of the state discrimination model training method can obtain a large number of training samples.

[0021] The execution subject of the state discrimination model training method is usually a server. The server can be hardware or software. When the server is hardware, it can be implemented as a distributed server cluster composed of multiple servers, or as a single server. When the server is software, it can be implemented as multiple software or software modules (for example, to provide distributed services), or as a single software or software module. No specific limitation is made here.

[0022] Generally, a map navigation application is installed on the terminal of the sample user. The sample user can query routes through the map navigation application and travel under the route guidance. At the same time, the map navigation application can also record the sample navigation data generated by the sample user during the travel. The sample navigation data is preprocessed, and the information of the sample user's two adjacent navigation behaviors is obtained after filtering.

[0023] The training samples may include information on two adjacent navigation behaviors of a sample user and status labels of sample points of interest. The two adjacent navigation behaviors of the sample user may be two adjacent navigation behaviors of the sample user, including the navigation behavior of the sample user arriving at the first sample point of interest and the navigation behavior of the sample user arriving at the second sample point of interest from the first sample point of interest. The information on the navigation behavior may include, but is not limited to, basic information such as the category to which the point of interest belongs, the visiting distance, and the visiting duration. The status of the point of interest may include a normal status and an abnormal status. The normal status may be an open status. The abnormal status may include, but is not limited to, suspension of business, relocation, name change, and demolition, etc. By annotating the status of the first sample point of interest, the status label of the sample point of interest can be obtained. For example, the label value of an abnormal first sample point of interest may be 1, and the label value of a normal first sample point of interest may be 0.

[0024] Step 102: Analyze the information on two adjacent navigation behaviors of the sample user to determine the sample navigation behavior time series data.

[0025] In this embodiment, the above-mentioned execution entity may analyze the information on two adjacent navigation behaviors of the sample user to determine the sample navigation behavior time series data. Among them, the sample navigation behavior time series data may include information on two adjacent navigation behaviors of multiple sample users over multiple days, and is divided into sample normal navigation behavior time series data and sample abnormal navigation behavior data.

[0026] In some embodiments, determine the category similarity between the first sample point of interest and the second sample point of interest, and the visiting duration of the first sample point of interest; based on the category similarity and the visiting duration, determine the sample normal navigation behavior time series data and the sample abnormal navigation behavior time series data.

[0027] Generally, if the categories of the first sample point of interest and the second sample point of interest are not similar, and the visiting duration is greater than or equal to the threshold of the visiting duration statistical value corresponding to the category of the first sample point of interest, the information on the two adjacent navigation behaviors of the corresponding sample user is determined as the sample normal navigation behavior time series data. If the categories of the first sample point of interest and the second sample point of interest are similar, and the visiting duration is less than the threshold of the visiting duration statistical value corresponding to the category of the first sample point of interest, the information on the two adjacent navigation behaviors of the corresponding sample user is determined as the sample abnormal navigation behavior time series data.

[0028] For example, if the first sample point of interest is an Internet cafe and the second sample point of interest is an Internet cafe or a gaming venue, then the categories of the first sample point of interest and the second sample point of interest are similar. Another example, if the first sample point of interest is an Internet cafe and the second sample point of interest is a bank, then the categories of the first sample point of interest and the second sample point of interest are not similar.

[0029] Points of interest (POIs) of different categories can correspond to different threshold values for visit duration statistics. By statistically analyzing the visit durations of normal POIs of the same category, the visit duration statistical value of the normal POIs of this category can be obtained. For example, the visit duration statistical value of a gas station is 30s - 300s. If the visit duration of a certain gas station is less than 30s, then the visit duration of this gas station is abnormal. Another example, the visit duration statistical value of an Internet cafe is 300s - 10800s. If the visit duration of a certain Internet cafe is greater than 300s, then the visit duration of this Internet cafe is normal.

[0030] Step 103: Based on the sample navigation behavior time series data, obtain the sample navigation behavior time series features.

[0031] In this embodiment, the above-mentioned execution entity can obtain the sample navigation behavior time series features based on the sample navigation behavior time series data.

[0032] Generally, by using multi-scale windows and periods to enhance and enrich the sample navigation behavior time series data of multiple sample users over multiple days, the recall ability for the POI status in the entire PV segment can be improved, especially for important scenarios such as daily high-timeliness recall and mid-long tail scenario recall. Among them, PV is the user visit frequency. Taking 12 days as an example, the entire PV segment can include the user visit frequency from the minimum of 1 day to the maximum of 12 days.

[0033] In some embodiments, use multiple scale windows and periods to perform feature aggregation on the sample normal navigation behavior time series data to obtain the sample normal navigation behavior time series features. Similarly, use multiple scale windows and periods to perform feature aggregation on the sample abnormal navigation behavior time series data to obtain the sample abnormal navigation behavior time series features. Integrate the sample normal navigation behavior time series features and the sample abnormal navigation behavior time series features to obtain the sample navigation behavior time series features.

[0034] For example, the scale window can range from a minimum of 1 day to a maximum of 12 days, and the period can be 30 days. Using a 1-day scale window and a 30-day period to perform feature aggregation on 30 days of sample navigation behavior time series data, a 30-dimensional array is obtained. Among them, the elements of the array are the values of the 1-day sample navigation behavior time series data. Using a 12-day scale window and a 30-day period to perform feature aggregation on 360 days of sample navigation behavior time series data, a 30-dimensional array is obtained. Among them, the elements of the array are the cumulative values of the 12-day sample navigation behavior time series data. Integrate the 30-dimensional arrays of 12 scale windows to obtain a 12×30 matrix.

[0035] Step 104: Input the sample navigation behavior time series features into the time series model to obtain the discrimination score of the sample navigation behavior time series features.

[0036] In this embodiment, the above-mentioned execution entity may input the temporal features of the sample navigation behavior into the temporal model to obtain the discrimination score of the temporal features of the sample navigation behavior.

[0037] On the basis of generating high-dimensional temporal features, advanced multi-temporal modeling methods can be used for feature processing, and quantitative analysis can be carried out for the performance of different temporal features to generate the discrimination score of the temporal features of the navigation behavior. The temporal models used may include but are not limited to: TimesNet, a time series prediction model based on Transformer, TCN (Temporal Convolutional Networks), etc.

[0038] TimesNet is a new deep learning model for time series analysis. Its core innovation lies in transforming the time series into the frequency domain to analyze periodicity, and capturing complex time patterns through multi-scale two-dimensional convolution, significantly improving the performance of prediction and classification tasks.

[0039] The time series prediction model based on Transformer adapts the successful Transformer architecture in NLP (Natural Language Processing) to the time series field, and is a deep learning model that captures long-term dependencies through the self-attention mechanism. Transformer significantly improves the ability to process long sequences through parallel computing and global information interaction.

[0040] TCN is a time series modeling method based on CNN (convolutional neural network), designed specifically for capturing long-term dependencies in sequence data. By replacing traditional recurrent neural networks with causal convolution and dilated convolution, it shows higher parallelism and better long-range dependency modeling ability in temporal tasks.

[0041] Through these models, deep processing of the temporal features is carried out to generate the discrimination score of the temporal features. Among them, the discrimination score can be used to measure the contribution degree of different temporal features to the target behavior, ensuring accurate perception of temporal changes in different scenarios.

[0042] Step 105, calculate the loss based on the discrimination score of the temporal features of the sample navigation behavior and the sample point-of-interest status label.

[0043] In this embodiment, the above-mentioned execution entity may input the discrimination score of the sample navigation behavior temporal feature and the sample interest point status label into the loss function to calculate the loss. The loss can be used to characterize the difference between the discrimination score and the sample interest point status label. The greater the difference, the greater the loss; the smaller the difference, the smaller the loss.

[0044] Step 106: Adjust the parameters of the temporal model based on the loss to obtain the status discrimination model.

[0045] In this embodiment, the above-mentioned execution entity may adjust the parameters of the temporal model based on the loss until the loss is minimized and the model converges to obtain the status discrimination model.

[0046] The embodiments of the present disclosure provide a method for training a status discrimination model. After the training and optimization of the status discrimination model are completed, inference calculation is performed. Quantitative analysis is carried out on the performance of different temporal features to generate the discrimination score of the navigation behavior temporal feature. This score measures the contribution degree of different navigation behavior temporal features to the target behavior, ensuring that the system can accurately perceive temporal changes in different scenarios.

[0047] Continue to refer to Figure 2 , which shows the flowchart 200 of another embodiment of the method for training a status discrimination model according to the present disclosure. The method for training the status discrimination model includes the following steps: Step 201: Obtain training samples.

[0048] In this embodiment, the specific operation of step 201 has been described in detail in step 101 of the embodiment shown in Figure 1 , and will not be elaborated here.

[0049] Step 202: Match the categories of the sample's first interest point and the sample's second interest point in the category similarity matching table to obtain the first matching result.

[0050] In this embodiment, the execution entity of the method for training the status discrimination model may match the categories of the sample's first interest point and the sample's second interest point in the category similarity matching table to obtain the first matching result.

[0051] The category similarity matching table may be generated by statistically analyzing the categories of the abnormal first interest point and the second interest point in the historical adjacent two navigation behaviors. Generally, the information of the historical adjacent two navigation behaviors is obtained, the historical adjacent two navigation behaviors containing the abnormal first interest point are filtered out, and the categories of the abnormal first interest point and the second interest point in these historical adjacent two navigation behaviors are statistically analyzed to generate the category similarity matching table. Each piece of information in the category similarity matching table may include the category of the abnormal first interest point and the category of the second interest point in the historical adjacent two navigation behaviors.

[0052] Step 203: Determine the category similarity between the sample first interest point and the sample second interest point according to the first matching result.

[0053] In this embodiment, the above execution entity may determine the category similarity between the sample first interest point and the sample second interest point according to the first matching result.

[0054] Generally, if the categories of the sample first interest point and the sample second interest point match a piece of information in the category similarity matching table, the sample first interest point and the sample second interest point are category-similar. If the categories of the sample first interest point and the sample second interest point do not match each piece of information in the category similarity matching table, the sample first interest point and the sample second interest point are not category-similar.

[0055] Step 204: Match the visit duration in the normal visit duration statistical table to obtain a second matching result.

[0056] In this embodiment, the above execution entity may match the visit duration in the normal visit duration statistical table to obtain a second matching result.

[0057] The normal visit duration statistical table may be generated by statistically counting the visit durations of normal interest points of various categories, including the visit duration statistical values of normal interest points of various categories. For example, the visit duration statistical value of a normal gas station is 30s - 300s. The visit duration statistical value of a normal Internet cafe is 300s - 10800s.

[0058] Step 205: Determine the sample normal navigation behavior time series data and the sample abnormal navigation behavior time series data according to the second matching result and the category similarity.

[0059] In this embodiment, the above execution entity may determine the sample normal navigation behavior time series data and the sample abnormal navigation behavior time series data according to the second matching result and the category similarity.

[0060] Generally, if the categories of the sample first interest point and the sample second interest point are not similar, and the visit duration is greater than or equal to the visit duration statistical value threshold corresponding to the category of the sample first interest point, the information of two adjacent navigation behaviors of the corresponding sample user is determined as the sample normal navigation behavior time series data. If the categories of the sample first interest point and the sample second interest point are similar, and the visit duration is less than the visit duration statistical value threshold corresponding to the category of the sample first interest point, the information of two adjacent navigation behaviors of the corresponding sample user is determined as the sample abnormal navigation behavior time series data.

[0061] For example, if the first sample point of interest is an Internet café and the second sample point of interest is an Internet café or a gaming venue, then the categories of the first and second sample points of interest are similar. Another example is that if the first sample point of interest is an Internet café and the second sample point of interest is a bank, then the categories of the first and second sample points of interest are not similar.

[0062] Points of interest of different categories can correspond to different threshold values for the statistical value of the visit duration. By statistically analyzing the visit duration of normal points of interest of the same category, the statistical value of the visit duration of the normal points of interest of this category can be obtained. For example, the statistical value of the visit duration of a gas station is 30s - 300s. If the visit duration of a certain gas station is less than 30s, then the visit duration of this gas station is abnormal. Another example is that the statistical value of the visit duration of an Internet café is 300s - 10800s. If the visit duration of a certain Internet café is greater than 300s, then the visit duration of this Internet café is normal.

[0063] Step 206: Use multiple scale windows and periods to perform feature aggregation on the sample normal navigation behavior time series data to obtain the sample normal navigation behavior time series features.

[0064] In this embodiment, the above execution entity can use multiple scale windows and periods to perform feature aggregation on the sample normal navigation behavior time series data to obtain the sample normal navigation behavior time series features.

[0065] Generally, by using multiple scale windows and periods to enhance and enrich the sample normal navigation behavior time series data of multiple sample users over multiple days, the recall ability for the POI status of the entire PV segment can be improved, especially for important scenarios such as daily high-timeliness recall and medium and long-tail scenario recall. Among them, PV is the user visit frequency. Taking 12 days as an example, the entire PV segment can include the user visit frequency from the minimum of 1 day to the maximum of 12 days.

[0066] For example, the scale window can range from a minimum of 1 day to a maximum of 12 days, and the period can be 30 days. Using a 1-day scale window and a 30-day period to perform feature aggregation on 30 days of sample normal navigation behavior time series data, a 30-dimensional array is obtained. Among them, the elements of the array are the values of the 1-day sample normal navigation behavior time series data. Using a 12-day scale window and a 30-day period to perform feature aggregation on 360 days of sample normal navigation behavior time series data, a 30-dimensional array is obtained. Among them, the elements of the array are the cumulative values of the 12-day sample normal navigation behavior time series data. Integrating the 30-dimensional arrays of 12 scale windows results in a 12×30 matrix.

[0067] Step 207: Use multiple scale windows and periods to perform feature aggregation on the sample abnormal navigation behavior time series data to obtain the sample abnormal navigation behavior time series features.

[0068] In this embodiment, the above-mentioned execution entity can use multiple scale windows and periods to perform feature aggregation on the time series data of sample abnormal navigation behaviors, and obtain the time series features of sample abnormal navigation behaviors.

[0069] Generally, by using multiple scale windows and periods to enhance and enrich the time series data of sample abnormal navigation behaviors of multiple sample users over multiple days, the recall ability for the POI status of the entire PV segment can be improved, especially for important scenarios such as daily high-timeliness recall and medium-long tail scenario recall. Among them, PV is the user visit frequency. Taking 12 days as an example, the entire PV segment can include the user visit frequency from the minimum of 1 day to the maximum of 12 days.

[0070] For example, the scale window can range from a minimum of 1 day to a maximum of 12 days, and the period can be 30 days. Using a 1-day scale window and a 30-day period to perform feature aggregation on the time series data of 30 days of sample abnormal navigation behaviors, a 30-dimensional array is obtained. Among them, the elements of the array are the values of the time series data of 1-day sample abnormal navigation behaviors. Using a 12-day scale window and a 30-day period to perform feature aggregation on the time series data of 360 days of sample abnormal navigation behaviors, a 30-dimensional array is obtained. Among them, the elements of the array are the accumulated values of the time series data of 12-day sample abnormal navigation behaviors. Integrating the 30-dimensional arrays of 12 scale windows, a 12×30 matrix is obtained.

[0071] Step 208: Integrate the time series features of sample normal navigation behaviors and the time series features of sample abnormal navigation behaviors to obtain the time series features of sample navigation behaviors.

[0072] In this embodiment, the above-mentioned execution entity can integrate the time series features of sample normal navigation behaviors and the time series features of sample abnormal navigation behaviors to obtain the time series features of sample navigation behaviors.

[0073] For example, integrating the 12×30 time series features of sample normal navigation behaviors and the 12×30 time series features of sample abnormal navigation behaviors, a 12×60 matrix is obtained.

[0074] Step 209: Input the time series features of sample navigation behaviors into the time series model to obtain the discrimination scores of the time series features of sample navigation behaviors.

[0075] Step 210: Calculate the loss based on the discrimination scores of the time series features of sample navigation behaviors and the sample point-of-interest status labels.

[0076] Step 211: Adjust the parameters of the time series model based on the loss to obtain the status discrimination model.

[0077] In this embodiment, the specific operations of steps 209-211 have been described in Figure 1The steps 104-106 in the illustrated embodiment are introduced in detail and will not be elaborated here.

[0078] The embodiment of the present disclosure provides a method for training a state discrimination model. After the training and optimization of the state discrimination model are completed, inference calculation is performed. Quantitative analysis is carried out for the performances of different temporal features to generate discrimination scores for the temporal features of navigation behaviors. This score measures the contribution degree of different temporal features of navigation behaviors to the target behavior, ensuring that the system can accurately perceive temporal changes in different scenarios.

[0079] Further referring to Figure 3 , which shows a flow 300 of an embodiment of the method for identifying the state of a point of interest according to the present disclosure. The method for identifying the state of a point of interest includes the following steps: Step 301, obtaining information on two adjacent navigation behaviors.

[0080] In this embodiment, the execution subject of the method for identifying the state of a point of interest can obtain information on two adjacent navigation behaviors.

[0081] The execution subject of the method for identifying a point of interest is usually a server. The server can be hardware or software. When the server is hardware, it can be implemented as a distributed server cluster composed of multiple servers or as a single server. When the server is software, it can be implemented as multiple software or software modules (for example, used to provide distributed services) or as a single software or software module. No specific limitation is made here.

[0082] Generally, a map navigation application is installed on the user's terminal. The user can query routes through the map navigation application and travel under the route guidance. At the same time, the map navigation application can also record the navigation data generated by the user during the travel. By preprocessing the navigation data, the information on two adjacent navigation behaviors can be filtered out.

[0083] Two adjacent navigation behaviors can be two adjacent navigation behaviors of the user, including the navigation behavior of the user arriving at the first point of interest and the navigation behavior of the user arriving at the second point of interest from the first point of interest. The information on the navigation behavior can include but is not limited to basic information such as the category to which the point of interest belongs, the arrival distance, and the arrival duration. The state of the point of interest can include a normal state and an abnormal state. The normal state can be an open state. The abnormal state can include but is not limited to: closing, relocating, renaming, and demolishing, etc.

[0084] Step 302, analyzing the information on two adjacent navigation behaviors to determine the temporal data of the navigation behavior.

[0085] In this embodiment, the above-mentioned execution entity can analyze the information of two adjacent navigation behaviors to determine the navigation behavior time series data. Among them, the navigation behavior time series data can include the information of two adjacent navigation behaviors of multiple users over multiple days, and is divided into normal navigation behavior time series data and abnormal navigation behavior data.

[0086] In some embodiments, determine the category similarity between the first point of interest and the second point of interest, and the visit duration of the first point of interest; based on the category similarity and the visit duration, determine the normal navigation behavior time series data and the abnormal navigation behavior time series data.

[0087] Generally, if the categories of the first point of interest and the second point of interest are not similar, and the visit duration is greater than or equal to the threshold of the visit duration statistical value corresponding to the category of the first point of interest, the information of the two adjacent navigation behaviors corresponding thereto is determined as normal navigation behavior time series data. If the categories of the first point of interest and the second point of interest are similar, and the visit duration is less than the threshold of the visit duration statistical value corresponding to the category of the first point of interest, the information of the two adjacent navigation behaviors corresponding thereto is determined as abnormal navigation behavior time series data.

[0088] For example, if the first point of interest is an Internet cafe and the second point of interest is an Internet cafe or a gaming venue, then the categories of the first point of interest and the second point of interest are similar. For another example, if the first point of interest is an Internet cafe and the second point of interest is a bank, then the categories of the first point of interest and the second point of interest are not similar.

[0089] Points of interest of different categories can correspond to different thresholds of visit duration statistical values. By statistically analyzing the visit durations of normal points of interest of the same category, the visit duration statistical values of the normal points of interest of this category can be obtained. For example, the visit duration statistical value of a gas station is 30s - 300s. If the visit duration of a certain gas station is less than 30s, then the visit duration of this gas station is abnormal. For another example, the visit duration statistical value of an Internet cafe is 300s - 10800s. If the visit duration of a certain Internet cafe is greater than 300s, then the visit duration of this Internet cafe is normal.

[0090] Step 303, based on the navigation behavior time series data, obtain navigation behavior time series features.

[0091] In this embodiment, the above-mentioned execution entity can obtain navigation behavior time series features based on the navigation behavior time series data.

[0092] Generally, by enhancing and enriching the navigation behavior time series data of multiple users over multiple days through multi-scale windows and periods, the recall ability for the POI status of the entire PV segment can be improved, especially for important scenarios such as daily high-timeliness recall and medium and long-tail scenario recall. Among them, PV is the user visit frequency. Taking 12 days as an example, the entire PV segment can include the user visit frequency from the minimum of 1 day to the maximum of 12 days.

[0093] In some embodiments, multiple scale windows and periods are used to aggregate the temporal data of normal navigation behavior to obtain the temporal features of normal navigation behavior. Similarly, multiple scale windows and periods are used to aggregate the temporal data of abnormal navigation behavior to obtain the temporal features of abnormal navigation behavior. The temporal features of normal navigation behavior and the temporal features of abnormal navigation behavior are integrated to obtain the temporal features of navigation behavior.

[0094] For example, the scale window can range from a minimum of 1 day to a maximum of 12 days, and the period can be 30 days. Aggregating the temporal data of 30-day navigation behavior using a 1-day scale window and a 30-day period results in a 30-dimensional array. Among them, the elements of the array are the values of the 1-day navigation behavior temporal data. Aggregating the temporal data of 360-day navigation behavior using a 12-day scale window and a 30-day period results in a 30-dimensional array. Among them, the elements of the array are the accumulated values of the 12-day navigation behavior temporal data. Integrating the 30-dimensional arrays of 12 scale windows results in a 12×30 matrix.

[0095] Step 304, input the temporal features of navigation behavior into a pre-trained state discrimination model to obtain the discrimination score of the temporal features of navigation behavior.

[0096] In this embodiment, the above execution entity can input the temporal features of navigation behavior into a pre-trained state discrimination model to obtain the discrimination score of the temporal features of navigation behavior.

[0097] Based on the generated high-dimensional temporal features, advanced multi-temporal modeling methods can be used for feature processing, and quantitative analysis can be performed on the performance of different temporal features to generate the discrimination score of the temporal features of navigation behavior. The temporal models used can include but are not limited to: TimesNet, Transformer-based time series prediction models, TCN, etc.

[0098] TimesNet is a new deep learning model for time series analysis. Its core innovation lies in converting time series to the frequency domain to analyze periodicity and capturing complex time patterns through multi-scale two-dimensional convolution, significantly improving the performance of prediction and classification tasks.

[0099] The Transformer-based time series prediction model adapts the successful Transformer architecture in NLP to the time series field and is a deep learning model that captures long-term dependencies through the self-attention mechanism. Transformer significantly improves the ability to process long sequences through parallel computing and global information interaction.

[0100] TCN is a CNN-based time series modeling method designed specifically to capture long-term dependencies in sequential data. By replacing traditional recurrent neural networks with causal and dilated convolutions, it exhibits higher parallelism and better long-range dependency modeling capabilities in time series tasks.

[0101] These models perform in-depth processing on time series features to generate discriminant scores for the time series features. Among them, the discriminant scores can be used to measure the contribution degree of different time series features to the target behavior, ensuring accurate perception of time series changes in different scenarios.

[0102] Step 305: Determine the state of the first point of interest based on the discriminant score of the navigation behavior time series features.

[0103] In this embodiment, the above-mentioned execution entity can determine the state of the first point of interest based on the discriminant score of the navigation behavior time series features.

[0104] Generally, if the discriminant score of the navigation behavior time series features is greater than the first preset threshold, it is determined that the state of the first point of interest is an abnormal state; if the discriminant score of the navigation behavior time series features is less than or equal to the first preset threshold, it is determined that the state of the first point of interest is a normal state.

[0105] For points of interest in different vertical categories, the accuracy of using the discriminant score of the navigation behavior time series features to determine the state of the first point of interest is different.

[0106] For points of interest with a strong correlation with the user's navigation behavior characteristics (points of interest that the vast majority of users choose to reach under the guidance of navigation, such as banks, business halls, gas stations, auto services, etc.), usually only using the discriminant score of the navigation behavior time series features can determine the state of the first point of interest with relatively high accuracy.

[0107] For points of interest with a weak correlation with the user's navigation behavior characteristics (points of interest that only some users choose to reach under the guidance of navigation, such as stores, companies, residential communities, etc.), the accuracy of using only the discriminant score of the navigation behavior time series features to determine the state of the first point of interest is relatively low. To further improve the accuracy and reliability of the data, multivariate falsification information can be combined to calibrate the discriminant score of the navigation behavior time series features. Specifically, obtain the falsification information of the first point of interest; perform weighted summation on the falsification information and the discriminant score to obtain the weighted summation result; determine the state of the first point of interest based on the weighted summation result. If the weighted summation result is greater than the second preset threshold, it is determined that the state of the first point of interest is an abnormal state; if the weighted summation result is less than or equal to the second preset threshold, it is determined that the state of the first point of interest is a normal state.

[0108] Among them, the falsification information may include, but is not limited to, at least one of the following: location data, stay point data, etc. Generally, for a point of interest with a weak correlation with the user's navigation behavior characteristics, the more types of falsification information combined, the higher the accuracy of the determined status of the first point of interest. For example, the accuracy of the status of a point of interest determined by only combining location data or stay point data is lower than that determined by combining both location data and stay point data.

[0109] The location data can be, for example, WiFi location data. For example, for a point of interest, users visit it every day, and the users who visit will connect to the WiFi of this point of interest. Based on the number of users connecting to the WiFi every day, the first decline score can be calculated. The first decline score can be used to characterize the decline of the number of users connecting to the WiFi over time.

[0110] The stay point data can be the data of the user's stay at the point of interest. For example, for a point of interest, users visit it every day, and the users who visit will stay for different lengths of time. The number of users whose stay time is greater than a preset time threshold (such as 30 seconds) every day is counted as the number of stay points every day. Based on the number of stay points every day, the second decline score can be calculated. The second decline score can be used to characterize the decline of the number of stay points over time.

[0111] For a situation where a higher accuracy of the point of interest status is required, the weights of the discrimination score of the navigation behavior time series characteristics, the first decline score, and the second decline score can be obtained, and the discrimination score, the first decline score, and the second decline score are weighted and summed, and based on the weighted sum result, the status of the first point of interest is determined. Among them, the weights of the discrimination score, the first decline score, and the second decline score can be values set according to experience after investigating and evaluating a large number of points of interest with a weak correlation with the user's navigation behavior characteristics.

[0112] For a situation where a lower accuracy of the point of interest status is required, the weights of the discrimination score of the navigation behavior time series characteristics and the first decline score can be obtained, and the discrimination score and the first decline score are weighted and summed, and based on the weighted sum result, the status of the first point of interest is determined. Among them, the weights of the discrimination score and the first decline score can be values set according to experience after investigating and evaluating a large number of points of interest with a weak correlation with the user's navigation behavior characteristics. Or, the weights of the discrimination score of the navigation behavior time series characteristics and the second decline score can be obtained, and the discrimination score and the second decline score are weighted and summed, and based on the weighted sum result, the status of the first point of interest is determined. Among them, the weights of the discrimination score and the second decline score can be values set according to experience after investigating and evaluating a large number of points of interest with a weak correlation with the user's navigation behavior characteristics.

[0113] An embodiment of the present disclosure provides a method for identifying the status of a POI based on navigation behavior. By deeply utilizing navigation behavior, the type and residence duration of the POI, combined with excellent feature engineering and a superior deep multi-temporal network, the mining efficiency of important scenarios is improved in status mining, and the mining coverage of weak image scenarios is greatly enriched.

[0114] Further referring to Figure 4 , which shows a flow 400 of another embodiment of the method for identifying the status of an interest point according to the present disclosure. The method for identifying the status of an interest point includes the following steps: Step 401, obtaining information on two adjacent navigation behaviors.

[0115] In this embodiment, the specific operation of step 401 has been introduced in detail in step 301 of the embodiment shown in Figure 3 , and will not be elaborated here.

[0116] Step 402, matching the category of the first interest point and the category of the second interest point in the category similarity matching table to obtain a first matching result.

[0117] In this embodiment, the execution subject of the method for identifying the status of an interest point may match the category of the first interest point and the category of the second interest point in the category similarity matching table to obtain a first matching result.

[0118] The category similarity matching table may be generated by statistically analyzing the categories of the abnormal first interest point and the second interest point in the historical two adjacent navigation behaviors. Generally, information on the historical two adjacent navigation behaviors is obtained, the historical two adjacent navigation behaviors containing the abnormal first interest point are filtered out, and the categories of the abnormal first interest point and the second interest point in these historical two adjacent navigation behaviors are statistically analyzed to generate the category similarity matching table. Each piece of information in the category similarity matching table may include the category of the abnormal first interest point and the category of the second interest point in the two adjacent navigation behaviors.

[0119] Step 403, determining the category similarity between the first interest point and the second interest point according to the first matching result.

[0120] In this embodiment, the above-mentioned execution subject may determine the category similarity between the first interest point and the second interest point according to the first matching result.

[0121] Generally, if the category of the first interest point and the category of the second interest point match a piece of information in the category similarity matching table, the first interest point and the second interest point are similar in category. If the category of the first interest point and the category of the second interest point do not match any piece of information in the category similarity matching table, the first interest point and the second interest point are not similar in category.

[0122] Step 404: Match the visit duration in the normal visit duration statistical table to obtain a second matching result.

[0123] In this embodiment, the above-mentioned execution entity can match the visit duration in the normal visit duration statistical table to obtain a second matching result.

[0124] The normal visit duration statistical table can be generated by statistically calculating the visit durations of various types of normal points of interest, including the visit duration statistical values of various types of normal points of interest. For example, the visit duration statistical value of a normal gas station is 30s - 300s. The visit duration statistical value of a normal Internet cafe is 300s - 10,800s.

[0125] Step 405: Determine the normal navigation behavior time series data and abnormal navigation behavior time series data according to the second matching result and category similarity.

[0126] In this embodiment, the above-mentioned execution entity can determine the normal navigation behavior time series data and abnormal navigation behavior time series data according to the second matching result and category similarity.

[0127] Generally, if the categories of the first point of interest and the second point of interest are not similar, and the visit duration is greater than or equal to the visit duration statistical value threshold corresponding to the category of the first point of interest, the information of the corresponding two adjacent navigation behaviors is determined as the normal navigation behavior time series data. If the categories of the first point of interest and the second point of interest are similar, and the visit duration is less than the visit duration statistical value threshold corresponding to the category of the first point of interest, the information of the corresponding two adjacent navigation behaviors is determined as the abnormal navigation behavior time series data.

[0128] For example, if the first point of interest is an Internet cafe and the second point of interest is an Internet cafe or a game venue, then the categories of the first point of interest and the second point of interest are similar. For another example, if the first point of interest is an Internet cafe and the second point of interest is a bank, then the categories of the first point of interest and the second point of interest are not similar.

[0129] Different categories of points of interest can correspond to different visit duration statistical value thresholds. By statistically calculating the visit durations of normal points of interest in the same category, the visit duration statistical value of the normal points of interest in this category can be obtained. For example, the visit duration statistical value of a gas station is 30s - 300s. If the visit duration of a certain gas station is less than 30s, then the visit duration of this gas station is abnormal. For another example, the visit duration statistical value of an Internet cafe is 300s - 10,800s. If the visit duration of a certain Internet cafe is greater than 300s, then the visit duration of this Internet cafe is normal.

[0130] Step 406: Use multiple scale windows and periods to perform feature aggregation on the normal navigation behavior time series data to obtain normal navigation behavior time series features.

[0131] In this embodiment, the above-mentioned execution entity can use multiple scale windows and periods to perform feature aggregation on the time series data of normal navigation behaviors, so as to obtain the time series features of normal navigation behaviors.

[0132] Generally, by using multiple scale windows and periods to enhance and enrich the time series data of normal navigation behaviors of multiple users over multiple days, the recall ability for the POI status of the entire PV segment can be improved, especially for important scenarios such as daily high-timeliness recall and mid-long tail scenario recall. Among them, PV is the user visit frequency. Taking 12 days as an example, the entire PV segment can include the user visit frequency from the minimum of 1 day to the maximum of 12 days.

[0133] For example, the scale window can range from a minimum of 1 day to a maximum of 12 days, and the period can be 30 days. Using a 1-day scale window and a 30-day period to perform feature aggregation on 30 days of time series data of normal navigation behaviors, a 30-dimensional array is obtained. Among them, the elements of the array are the values of the time series data of normal navigation behaviors for 1 day. Using a 12-day scale window and a 30-day period to perform feature aggregation on 360 days of time series data of normal navigation behaviors, a 30-dimensional array is obtained. Among them, the elements of the array are the accumulated values of the time series data of normal navigation behaviors for 12 days. Integrating the 30-dimensional arrays of 12 scale windows, a 12×30 matrix is obtained.

[0134] Step 407, use multiple scale windows and periods to perform feature aggregation on the time series data of abnormal navigation behaviors, so as to obtain the time series features of abnormal navigation behaviors.

[0135] In this embodiment, the above-mentioned execution entity can use multiple scale windows and periods to perform feature aggregation on the time series data of abnormal navigation behaviors, so as to obtain the time series features of abnormal navigation behaviors.

[0136] Generally, by using multiple scale windows and periods to enhance and enrich the time series data of abnormal navigation behaviors of multiple users over multiple days, the recall ability for the POI status of the entire PV segment can be improved, especially for important scenarios such as daily high-timeliness recall and mid-long tail scenario recall. Among them, PV is the user visit frequency. Taking 12 days as an example, the entire PV segment can include the user visit frequency from the minimum of 1 day to the maximum of 12 days.

[0137] For example, the scale window can range from a minimum of 1 day to a maximum of 12 days, and the period can be 30 days. Aggregating the temporal data of abnormal navigation behavior for 30 days using a 1-day scale window and a 30-day period results in a 30-dimensional array. Among them, the elements of the array are the values of the temporal data of abnormal navigation behavior for 1 day. Aggregating the temporal data of abnormal navigation behavior for 360 days using a 12-day scale window and a 30-day period results in a 30-dimensional array. Among them, the elements of the array are the accumulated values of the temporal data of abnormal navigation behavior for 12 days. Integrating the 30-dimensional arrays of 12 scale windows results in a 12×30 matrix.

[0138] Step 408: Integrate the temporal features of normal navigation behavior and the temporal features of abnormal navigation behavior to obtain the temporal features of navigation behavior.

[0139] In this embodiment, the above-mentioned execution subject can integrate the temporal features of normal navigation behavior and the temporal features of abnormal navigation behavior to obtain the temporal features of navigation behavior.

[0140] For example, integrating the 12×30 temporal features of normal navigation behavior and the 12×30 temporal features of abnormal navigation behavior results in a 12×60 matrix.

[0141] Step 409: Input the temporal features of navigation behavior into a pre-trained state discrimination model to obtain the discrimination score of the temporal features of navigation behavior.

[0142] Step 410: Determine the state of the first point of interest based on the discrimination score of the temporal features of navigation behavior.

[0143] In this embodiment, the specific operations of steps 409-410 have been introduced in detail in steps 304-305 of the embodiment shown in Figure 3 and will not be elaborated here.

[0144] The present disclosure embodiment provides a method for identifying the state of a POI based on navigation behavior. By deeply utilizing navigation behavior, the type and residence duration of the POI, combining excellent feature engineering and a good deep multi-temporal network, the mining efficiency of important scenarios is improved in state mining, and the mining coverage of weak image scenarios is greatly enriched.

[0145] The technical solution of the method for identifying the state of a POI based on navigation behavior can be divided into three main modules: a navigation behavior acquisition module, a feature extraction module, and a multi-temporal model mining module, aiming to accurately capture and model the temporal features of navigation behavior to improve the accuracy and robustness of navigation behavior analysis.

[0146] Figure 5 Shows a schematic diagram of the navigation behavior acquisition module.

[0147] The core objective of the navigation behavior acquisition module is to extract the basic sequential data of navigation behavior, providing data support for subsequent feature analysis. The specific process is as follows: 1. Navigation data collation: Preprocess the navigation data to obtain basic information such as the category of the POI for adjacent two navigation behaviors, the arrival distance, and the arrival duration.

[0148] 2. Vertical category similarity matching and abnormal behavior marking: By counting the navigation behaviors reaching expired POIs, organize the mapping relationships of abnormal navigation behaviors for each vertical category, construct a vertical category similarity matching table, map the relationships of POI vertical categories of the same type to facilitate the discrimination of abnormal navigation behaviors. Combining with real navigation data, count the arrival durations of different vertical categories and formulate the 95th percentile statistical value, and mark the behaviors with abnormal arrival durations (below the statistical threshold) as abnormal.

[0149] 3. Construction of long-term navigation behavior sequential information: Generate long-term navigation behavior sequential information based on the abnormal discrimination results for subsequent sequential modeling analysis.

[0150] Figure 6 The schematic diagram of the feature extraction module is shown.

[0151] The feature extraction module enhances and enriches the sequential features of navigation behavior through multi-scale windows, aiming to improve the recall ability of the algorithm for the POI status in the entire PV segment, including: high-timeliness recall for important scenarios at the daily level and recall improvement for medium and long-tail scenarios. The specific process is as follows: 1. Aggregation of sequential features of normal navigation behavior: Aggregate the sequential data of normal navigation behavior using windows and periods of different scales from 1 day to 12 days.

[0152] 2. Aggregation of sequential features of abnormal navigation behavior: Aggregate the sequential data of abnormal navigation behavior using windows and periods of different scales from 1 day to 12 days.

[0153] 3. Integration of multi-dimensional features: Integrate the obtained multi-dimensional sequential features of normal navigation behavior and abnormal navigation behavior.

[0154] Figure 7 The schematic diagram of the multi-sequential model mining module is shown.

[0155] The multi-sequential model mining module conducts in-depth analysis and discrimination on multi-dimensional sequential features through multi-sequential models to achieve accurate modeling of sequential change patterns. Its core objective is to effectively capture complex periodic change features and improve the discrimination ability of the model and the accuracy of the overall algorithm by deeply integrating the internal correlation between multi-dimensional sequential features. The specific process is as follows: 1. Multi-temporal Model Construction and Feature Processing: Based on the high-dimensional temporal features generated by the feature extraction module, advanced multi-temporal modeling methods are used for feature processing. The temporal models used include but are not limited to: TimesNet, Transformer-based time series prediction models, TCN, etc. Through these models, in-depth pattern learning and feature extraction are performed on the temporal data, thereby generating discriminant scores for multi-temporal features, laying a foundation for subsequent analysis.

[0156] 2. Model Inference and Discriminant Score Calculation: After completing the training and optimization of the multi-temporal model, inference calculations are performed, and quantitative analysis is carried out on the performance of different temporal features to generate discriminant scores for navigation behavior features. This score measures the contribution degree of different temporal features to the target behavior, ensuring that the system can accurately perceive temporal changes in different scenarios.

[0157] 3. Data Calibration: Combining multi-source verification information such as WiFi positioning data and stay point data, the discriminant results are calibrated to further improve the accuracy and reliability of the data.

[0158] For further reference Figure 8 , as an implementation of the methods shown in the above figures, the present disclosure provides an embodiment of a state discrimination model training device. This device embodiment corresponds to Figure 1 the method embodiment shown, and this device can be specifically applied to various electronic devices.

[0159] As Figure 8 shown, the state discrimination model training device 800 of this embodiment may include: a first acquisition module 801, an analysis module 802, a second acquisition module 803, a discrimination module 804, a calculation module 805, and an adjustment module 806. Among them, the first acquisition module 801 is configured to acquire training samples, where the training samples include information on two adjacent navigation behaviors of a sample user and a sample interest point status label. The two adjacent navigation behaviors of the sample user include the navigation behavior of the sample user reaching the sample first interest point and the navigation behavior of reaching the sample second interest point from the sample first interest point, and the sample interest point status label is the status of the sample first interest point; the analysis module 802 is configured to analyze the information on two adjacent navigation behaviors of the sample user to determine the sample navigation behavior temporal data; the second acquisition module 803 is configured to acquire sample navigation behavior temporal features based on the sample navigation behavior temporal data; the discrimination module 804 is configured to input the sample navigation behavior temporal features into the temporal model to obtain the discriminant score of the sample navigation behavior temporal features; the calculation module 805 is configured to calculate the loss based on the discriminant score of the sample navigation behavior temporal features and the sample interest point status label; the adjustment module 806 is configured to adjust the parameters of the temporal model based on the loss to obtain the state discrimination model.

[0160] In this embodiment, in the state discrimination model training device 800: For the specific processing of the first acquisition module 801, the analysis module 802, the second acquisition module 803, the discrimination module 804, the calculation module 805, and the adjustment module 806 and the technical effects brought thereby, reference can be made respectively to Figure 1 the relevant descriptions of steps 101-106 in the corresponding embodiment, which will not be elaborated here.

[0161] In some optional implementation manners of this embodiment, the analysis module 802 includes: a first determination sub-module configured to determine the category similarity between the sample first interest point and the sample second interest point, and the visit duration of the sample first interest point; a second determination sub-module configured to determine the sample normal navigation behavior time series data and the sample abnormal navigation behavior time series data based on the category similarity and the visit duration.

[0162] In some optional implementation manners of this embodiment, the first determination sub-module is further configured to: match the category of the sample first interest point and the category of the sample second interest point in the category similarity matching table to obtain a first matching result, where the category similarity matching table is generated by statistically analyzing the categories of the abnormal first interest point and the second interest point in two adjacent historical navigation behaviors; determine the category similarity between the sample first interest point and the sample second interest point according to the first matching result.

[0163] In some optional implementation manners of this embodiment, the second determination sub-module is further configured to: match the visit duration in the normal visit duration statistical table to obtain a second matching result, where the normal visit duration statistical table is generated by statistically analyzing the visit durations of normal interest points of various categories; determine the sample normal navigation behavior time series data and the sample abnormal navigation behavior time series data according to the second matching result and the category similarity.

[0164] In some optional implementation manners of this embodiment, the second determination sub-module is further configured to: if the categories of the sample first interest point and the sample second interest point are not similar, and the access duration is greater than or equal to the access duration statistical value threshold corresponding to the category of the sample first interest point, determine the information of two adjacent navigation behaviors of the corresponding sample user as the sample normal navigation behavior time series data; if the categories of the sample first interest point and the sample second interest point are similar, and the access duration is less than the access duration statistical value threshold corresponding to the category of the sample first interest point, determine the information of two adjacent navigation behaviors of the corresponding sample user as the sample abnormal navigation behavior time series data.

[0165] In some alternative implementation manners of this embodiment, the second acquisition module is further configured to: perform feature aggregation on the time series data of the sample normal navigation behavior using multiple scale windows and periods to obtain the time series features of the sample normal navigation behavior; perform feature aggregation on the time series data of the sample abnormal navigation behavior using multiple scale windows and periods to obtain the time series features of the sample abnormal navigation behavior; and integrate the time series features of the sample normal navigation behavior and the time series features of the sample abnormal navigation behavior to obtain the time series features of the sample navigation behavior.

[0166] Further referring to Figure 9 , as an implementation of the methods shown in the above figures, the present disclosure provides an embodiment of a point of interest status recognition device, and this device embodiment corresponds to Figure 3 the method embodiment shown, and this device can be specifically applied to various electronic devices.

[0167] As Figure 9 shown, the point of interest status recognition device 900 in this embodiment may include: a first acquisition module 901, an analysis module 902, a second acquisition module 903, a discrimination module 904, and a determination module 905. Among them, the first acquisition module 901 is configured to acquire information on two adjacent navigation behaviors, where the two adjacent navigation behaviors include the navigation behavior to reach the first point of interest and the navigation behavior from the first point of interest to the second point of interest; the analysis module 902 is configured to analyze the information on the two adjacent navigation behaviors to determine the navigation behavior time series data; the second acquisition module 903 is configured to acquire the navigation behavior time series features based on the navigation behavior time series data; the discrimination module 904 is configured to input the navigation behavior time series features into a pre-trained status discrimination model to obtain the discrimination score of the navigation behavior time series features, where the status discrimination model is used to discriminate the status of the point of interest; and the determination module 905 is configured to determine the status of the first point of interest based on the discrimination score of the navigation behavior time series features.

[0168] In this embodiment, in the point of interest status recognition device 900: the specific processing of the first acquisition module 901, the analysis module 902, the second acquisition module 903, the discrimination module 904, and the determination module 905 and the technical effects brought thereby can respectively refer to Figure 3 the relevant descriptions of steps 301-305 in the corresponding embodiment, and details are not described herein again.

[0169] In some alternative implementation manners of this embodiment, the analysis module 902 includes: a first determination sub-module configured to determine the category similarity between the first point of interest and the second point of interest, and the visit duration of the first point of interest; and a second determination sub-module configured to determine the normal navigation behavior time series data and the abnormal navigation behavior time series data based on the category similarity and the visit duration.

[0170] In some alternative implementation manners of this embodiment, the first determination sub-module is further configured to: match the category of the first point of interest and the category of the second point of interest in a category similarity matching table to obtain a first matching result, where the category similarity matching table is generated by statistically analyzing the categories of the abnormal first point of interest and the second point of interest in two adjacent historical navigation behaviors; determine the category similarity between the first point of interest and the second point of interest according to the first matching result.

[0171] In some alternative implementation manners of this embodiment, the second determination sub-module is further configured to: match the visit duration in a normal visit duration statistical table to obtain a second matching result, where the normal visit duration statistical table is generated by statistically analyzing the visit durations of normal points of interest of various categories; determine the normal navigation behavior time series data and the abnormal navigation behavior time series data according to the second matching result and the category similarity.

[0172] In some alternative implementation manners of this embodiment, the second determination sub-module is further configured to: if the categories of the first point of interest and the second point of interest are not similar, and the visit duration is greater than or equal to the threshold value of the visit duration statistical value corresponding to the category of the first point of interest, determine the information of two adjacent navigation behaviors as normal navigation behavior time series data; if the categories of the first point of interest and the second point of interest are similar, and the visit duration is less than the threshold value of the visit duration statistical value corresponding to the category of the first point of interest, determine the information of two adjacent navigation behaviors as abnormal navigation behavior time series data.

[0173] In some alternative implementation manners of this embodiment, the second acquisition module is further configured to: perform feature aggregation on the normal navigation behavior time series data using multiple scale windows and periods to obtain normal navigation behavior time series features; perform feature aggregation on the abnormal navigation behavior time series data using multiple scale windows and periods to obtain abnormal navigation behavior time series features; integrate the normal navigation behavior time series features and the abnormal navigation behavior time series features to obtain navigation behavior time series features.

[0174] In some alternative implementation manners of this embodiment, the determination module 905 is further configured to: obtain the verification information of the first point of interest, where the verification information includes at least one of the following: positioning data, stay point data; perform weighted summation on the verification information and the discrimination score to obtain a weighted summation result; determine the state of the first point of interest based on the weighted summation result.

[0175] In the technical solution of the present disclosure, the acquisition, storage, and application of the user's personal information involved all comply with the provisions of relevant laws and regulations and do not violate public order and good customs.

[0176] According to the embodiments of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0177] Figure 10 FIG. shows a schematic block diagram of an exemplary electronic device 1000 that can be used to implement embodiments of the present disclosure. The electronic device is intended to represent various forms of digital computers, such as, for example, laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as, for example, personal digital assistants, cellular telephones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely exemplary and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0178] As Figure 10 shown, the device 1000 includes a computing unit 1001 that can perform various appropriate actions and processes in accordance with a computer program stored in a read-only memory (ROM) 1002 or a computer program loaded from a storage unit 1008 into a random access memory (RAM) 1003. In the RAM 1003, various programs and data required for the operation of the device 1000 can also be stored. The computing unit 1001, the ROM 1002, and the RAM 1003 are connected to each other via a bus 1004. An input / output (I / O) interface 1005 is also connected to the bus 1004.

[0179] A plurality of components in the device 1000 are connected to the I / O interface 1005, including: an input unit 1006, such as a keyboard, a mouse, etc.; an output unit 1007, such as various types of displays, speakers, etc.; a storage unit 1008, such as a magnetic disk, an optical disk, etc.; and a communication unit 1009, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 1009 allows the device 1000 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0180] The computing unit 1001 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 1001 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 1001 executes the various methods and processes described above, such as the point of interest status recognition method. For example, in some embodiments, the point of interest status recognition method can be implemented as a computer software program that is tangibly contained in a machine-readable medium, such as the storage unit 1008. In some embodiments, part or all of the computer program can be loaded and / or installed onto the device 1000 via the ROM 1002 and / or the communication unit 1009. When the computer program is loaded into the RAM 1003 and executed by the computing unit 1001, one or more steps of the point of interest status recognition method described above can be executed. Alternatively, in other embodiments, the computing unit 1001 can be configured to execute the point of interest status recognition method by any other suitable means (e.g., by means of firmware).

[0181] The various embodiments of the systems and techniques described above in this document can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGA), application-specific integrated circuits (ASIC), application-specific standard products (ASSP), systems-on-a-chip (SOC), complex programmable logic devices (CPLD), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a dedicated or general-purpose programmable processor, and can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit the data and instructions to the storage system, the at least one input device, and the at least one output device.

[0182] The program code for implementing the methods of the present disclosure can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when the program codes are executed by the processor or controller, the functions / operations specified in the flowcharts and / or block diagrams are implemented. The program code can be executed entirely on the machine, partially on the machine, as an independent software package partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0183] In the context of this disclosure, a machine-readable medium can be a tangible medium that can contain or store a program for use by or in connection with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0184] For purposes of providing an interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the computer. Other kinds of devices can also be used to provide for interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic, speech, or tactile input).

[0185] The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer having a graphical user interface or a web browser through which the user can interact with an implementation of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of a communication network include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0186] A computer system can include a client and a server. The client and the server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. The server can be a cloud server, a server of a distributed system, or a server incorporating a blockchain.

[0187] It should be understood that the various forms of processes shown above can be used, with steps reordered, added or deleted. For example, the steps recited in this disclosure can be executed in parallel, sequentially, or in a different order, as long as the desired results of the technical solution provided in this disclosure can be achieved, and no limitation is imposed herein.

[0188] The above specific embodiments do not constitute a limitation on the protection scope of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub - combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of this disclosure shall be included within the protection scope of this disclosure.

Claims

1. A method for identifying a point of interest state, comprising: Acquiring information about two adjacent navigation behaviors, wherein the two adjacent navigation behaviors include a navigation behavior of reaching a first point of interest and a navigation behavior of reaching a second point of interest from the first point of interest; Analyzing information of the two adjacent navigation behaviors to determine navigation behavior time series data; Based on the navigation behavior time series data, obtaining navigation behavior time series characteristics; Inputting the navigation behavior time series feature into a pre-trained state discrimination model to obtain a discrimination score of the navigation behavior time series feature, wherein the state discrimination model is used to discriminate the state of the point of interest; Based on the discrimination score of the temporal feature of the navigation behavior, the state of the first point of interest is determined.

2. The method according to claim 1, wherein: The analyzing the information of the two adjacent navigation behaviors to determine the navigation behavior time series data includes: Determining the category similarity between the first point of interest and the second point of interest, and the visit duration of the first point of interest; Based on the category similarity and the visit duration, normal navigation behavior time series data and abnormal navigation behavior time series data are determined.

3. The method according to claim 2, wherein: The determining the category similarity between the first point of interest and the second point of interest includes: Matching the category of the first point of interest and the category of the second point of interest in a category similarity matching table to obtain a first matching result, wherein the category similarity matching table is statistically generated by the categories of the abnormal first point of interest and the second point of interest in two adjacent historical navigation behaviors; According to the first matching result, category similarity between the first point of interest and the second point of interest is determined.

4. The method according to claim 2, wherein: The determining of normal navigation behavior time series data and abnormal navigation behavior time series data based on the category similarity and the visit duration includes: Matching the visit duration in a normal visit duration statistical table to obtain a second matching result, wherein the normal visit duration statistical table is generated by statistically analyzing the visit durations of normal points of interest of various categories; Normal navigation behavior time series data and abnormal navigation behavior time series data are determined according to the second matching result and the category similarity.

5. The method according to claim 4, wherein: The determining, according to the second matching result and the category similarity, normal navigation behavior time series data and abnormal navigation behavior time series data comprises: If the categories of the first point of interest and the second point of interest are not similar, and the visit duration is greater than or equal to the visit duration statistical value threshold corresponding to the category of the first point of interest, the information of the two corresponding adjacent navigation behaviors is determined as normal navigation behavior time series data; If the categories of the first point of interest and the second point of interest are similar, and the visit duration is less than the visit duration statistical value threshold corresponding to the category of the first point of interest, the information of the corresponding two adjacent navigation behaviors is determined as abnormal navigation behavior time series data.

6. The method according to any one of claims 2 to 5, wherein: The obtaining of navigation behavior time series features based on the navigation behavior time series data includes: Using multiple scale windows and cycles to perform feature aggregation on the normal navigation behavior time series data to obtain normal navigation behavior time series features; Using multiple scale windows and cycles to perform feature aggregation on the abnormal navigation behavior time series data to obtain abnormal navigation behavior time series features; The normal navigation behavior time sequence feature and the abnormal navigation behavior time sequence feature are integrated to obtain the navigation behavior time sequence feature.

7. The method according to claim 1, wherein: The determining the state of the first point of interest based on the discrimination score of the temporal characteristics of the navigation behavior includes: Acquire verification information of the first point of interest, wherein the verification information includes at least one of the following: positioning data, resident point data: Performing weighted summation on the false verification information and the discrimination score to obtain a weighted summation result; Based on the weighted sum result, a state of the first point of interest is determined.

8. A state discrimination model training method, comprising: Acquire a training sample, wherein the training sample includes information of two adjacent navigation behaviors of a sample user and a sample point of interest status label, the two adjacent navigation behaviors of the sample user include a navigation behavior of the sample user arriving at a sample first point of interest and a navigation behavior of the sample user arriving at a sample second point of interest from the sample first point of interest, and the sample point of interest status label is the status of the sample first point of interest; Analyze the information of two adjacent navigation behaviors of the sample user to determine the sample navigation behavior time series data; Based on the sample navigation behavior time series data, obtaining the sample navigation behavior time series characteristics; Inputting the sample navigation behavior temporal features into a temporal model to obtain a discriminant score of the sample navigation behavior temporal features; Calculating the loss based on the discriminant score of the temporal feature of the sample navigation behavior and the state label of the sample point of interest; The parameters of the time series model are adjusted based on the loss to obtain a state discrimination model.

9. A device for identifying the state of a point of interest, comprising: A first acquisition module is configured to acquire information of two adjacent navigation behaviors, wherein the two adjacent navigation behaviors include a navigation behavior of reaching a first point of interest and a navigation behavior of reaching a second point of interest from the first point of interest; An analysis module is configured to analyze information of the two adjacent navigation behaviors to determine navigation behavior time series data; A second acquisition module is configured to acquire navigation behavior time series features based on the navigation behavior time series data; A discrimination module is configured to input the navigation behavior time series feature into a pre-trained state discrimination model to obtain a discrimination score of the navigation behavior time series feature, wherein the state discrimination model is used to discriminate the state of the point of interest; The determination module is configured to determine the state of the first point of interest based on the discrimination score of the temporal characteristics of the navigation behavior.

10. A state discrimination model training device, comprising: A first acquisition module is configured to acquire a training sample, wherein the training sample includes information of two adjacent navigation behaviors of a sample user and a sample point of interest status label, the two adjacent navigation behaviors of the sample user include a navigation behavior of the sample user to reach a sample first point of interest and a navigation behavior from the sample first point of interest to reach a sample second point of interest, and the sample point of interest status label is the status of the sample first point of interest; An analysis module is configured to analyze information of two adjacent navigation behaviors of the sample user to determine sample navigation behavior time series data; A second acquisition module is configured to acquire sample navigation behavior time series features based on the sample navigation behavior time series data; A discrimination module, configured to input the sample navigation behavior temporal features into a temporal model to obtain a discrimination score of the sample navigation behavior temporal features; A calculation module configured to calculate the loss based on the discrimination score of the temporal feature of the sample navigation behavior and the state label of the sample point of interest; The adjustment module is configured to adjust the parameters of the timing model based on the loss to obtain a state discrimination model.

11. An electronic device, comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-7 or 8.

12. A non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to cause the computer to execute the method of any one of claims 1-7 or 8.

13. A computer program product, comprising a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 7 or 8.

Images