Method and device for training a poi arrival location point scoring model, determining a navigation path

Abstract

Embodiments disclosed in the specification provide a method and device for training a POI arrival location point scoring model. Positive samples and negative samples are constructed based on real user historical trips. The positive samples include POI location points and actual arrival location points in a user historical trip. The negative samples include POI location points and non-actual arrival location points in a user historical trip. When the arrival location point scoring model is trained based on the positive sample set and the negative sample set, the sample features are used as the input of the model, and the corresponding model output is the score of the arrival location point in the sample.

Description

TECHNICAL FIELD

[0001] Embodiments of the present specification relate to the field of information technology, and in particular, to a method and device for training a POI arrival location point scoring model to determine a navigation path. BACKGROUND

[0002] When going out in daily life, a user can use a travel navigation service on the Internet. Generally, the travel navigation service determines a place (such as a community or a store) that the user wants to go to, and then plans a navigation path to provide to the user. The place that the user wants to go to is generally referred to as a user point of interest (POI).

[0003] Therefore, for the travel navigation service, how to find an arrival location point that is convenient for the user to enter a POI place is a technical problem that needs to be solved in the field. SUMMARY

[0004] Embodiments of the present specification provide a method and device for training a POI arrival location point scoring model to determine a navigation path, so as to find an arrival location point that is convenient for the user to enter a POI place.

[0005] According to a first aspect of embodiments of the present specification, a method for training an arrival location point scoring model is provided, comprising:

[0006] obtaining a positive sample set and a negative sample set; wherein the positive sample includes a POI location point and an actual arrival location point in a user's historical trip; and the negative sample includes a POI location point and a non-actual arrival location point in a user's historical trip;

[0007] training the arrival location point scoring model based on the positive sample set and the negative sample set; wherein a sample feature is used as a model input, and a corresponding model output is a score for the arrival location point in the sample.

[0008] According to a second aspect of embodiments of the present specification, a method for determining a navigation path is provided, applied to a travel navigation service, and the method comprises:

[0009] obtaining a POI location point specified by a user, and obtaining a plurality of candidate arrival location points corresponding to the POI location point;

[0010] for each candidate arrival location point, calling the arrival location point scoring model trained by the method of the first aspect to score the candidate arrival location point according to the POI location point and the candidate arrival location point;

[0011] determining a navigation path corresponding to the POI location point by taking the candidate arrival location point with the highest score as the end point of the navigation path.

[0012] According to a third aspect of the embodiments of the present specification, a computing device is provided, comprising a memory, a processor; the memory is configured to store computer instructions executable on the processor, and the processor is configured to implement the method of the first aspect or the second aspect when executing the computer instructions.

[0013] In the above technical solution, the positive samples include POI location points and actual arrival location points in a user historical trip, and the negative samples include POI location points and non-actual arrival location points in a user historical trip. When training the arrival location point scoring model based on the positive sample set and the negative sample set, the sample features are input into the model, and the corresponding model output is the score of the arrival location point in the sample. The arrival location point scoring model can be used to score a plurality of candidate arrival location points corresponding to a POI location point. The higher the score of a candidate arrival location point is, the more suitable the candidate arrival location point is as the end point of the navigation path. The candidate arrival location point with the highest score can be determined as the end point of the navigation path, and the navigation path is provided to the user.

[0014] Through the above technical solution, since the actual arrival point in the user historical trip is the arrival location point that the user personally selects to conveniently enter the POI site in practice, the arrival location point scoring model can be trained based on the real user historical trip. The score of the trained arrival location point scoring model is more accurate. Furthermore, the arrival location point scoring model can be used to mine the arrival location point that is convenient for the user to enter the POI site according to the POI location point. BRIEF DESCRIPTION OF DRAWINGS

[0015] Figure 1 An exemplary method for training a POI arrival location point scoring model is provided.

[0016] Figure 2 An exemplary training architecture diagram of a POI arrival location point scoring model is provided.

[0017] Figure 3 An exemplary method for determining a navigation path is provided.

[0018] Figure 4 is a structural diagram of a computer readable storage medium provided by the present disclosure.

[0019] Figure 5 is a structural diagram of a computing device provided by the present disclosure.

[0020] In the drawings, like or corresponding elements are denoted by like or corresponding reference numerals. The number of elements in the drawings is used for illustration only and does not limit the scope of the application, and any naming is only for differentiation and does not have any limiting meaning. DETAILED DESCRIPTION

[0021] In order to enable those skilled in the art to better understand the technical solutions in the specification, the technical solutions in the specification will be clearly and completely described below in conjunction with the drawings in the embodiments of the specification. Obviously, the described embodiments are only part of the embodiments of the specification, not all. Based on the embodiments in the specification, all other embodiments obtained by those skilled in the art without creative labor should belong to the scope of protection of the specification.

[0022] It should be noted that: in other embodiments, the steps of the corresponding method are not necessarily performed in the order shown and described in the specification. In some other embodiments, the steps included in the method can be more or less than described in the specification. In addition, a single step described in the specification may, in other embodiments, be divided into multiple steps for description; and multiple steps described in the specification may, in other embodiments, be combined into a single step for description.

[0023] The travel navigation service described in the present disclosure generally refers to an Internet service that plans a navigation path according to the travel needs of a user (the departure location point of the user and the POI site that the user wants to go to). For example, the user can install an application APP (such as XX map APP) of the travel navigation service on his own mobile device, and when the user has travel needs, he can open the application APP and input the name of the POI he wants to go to. The application APP determines the corresponding arrival location point and the departure location point currently occupied by the user, and then plans a navigation path for the user.

[0024] It is easy to understand that a certain location point (such as the central location point of the region) within the region of the POI site can be used to represent the location of the POI, which is generally referred to as the POI location point. The POI location point is usually not the location of the entrance of the POI site. If the POI location point is directly used as the end point of the navigation path, it may lead to the situation that the user is not convenient to enter the site after reaching the end point of the path.

[0025] In the scenario of the user driving, the above situation will be more common. If the POI location point is directly used as the end point of the navigation path, it not only may lead to the situation that the user is not convenient to enter the POI site after driving to the end point of the path, but also the end point of the path may not be suitable for the user to park, causing an embarrassing experience for the user driving.

[0026] One possible idea is that an artificial intelligence model can be trained to explore a more suitable arrival location point as the end point of a navigation path according to a POI location point in a user's travel demand. However, the training of a general artificial intelligence model depends on a large number of manually annotated samples, and it is necessary to manually annotate as many POI location points as possible with matching arrival location points, which not only consumes manpower and is inefficient, but also makes it difficult to respond to changes in POI site entrances in the real world (as can be easily understood, when the POI site entrance changes, the corresponding arrival location point suitable for the navigation trajectory will also change, and the model needs to be retrained based on the manually reannotated samples), resulting in low iteration efficiency of the artificial intelligence model.

[0027] Therefore, the technical solution provided by the present disclosure is to construct positive samples and negative samples based on real user historical travels, the positive samples including a POI location point and an actual arrival location point in a user historical travel, and the negative samples including a POI location point and a non-actual arrival location point in a user historical travel. When training an arrival location point scoring model based on the positive sample set and the negative sample set, the sample features are used as the input of the model, and the corresponding model output is the score of the arrival location point in the sample. As can be easily understood, the highest score (such as 1) is usually used as the label of the positive sample, and the lowest score (such as 0) is usually used as the label of the negative sample. The arrival location point scoring model can be used to score several candidate arrival location points corresponding to a POI location point, and the higher the score of a candidate arrival location point, the more suitable the candidate arrival location point is as the end point of a navigation path. The candidate arrival location point with the highest score can be used as the end point of a navigation path to determine the navigation path provided to the user.

[0028] In this way, it is not necessary to rely on artificial standard samples, but to construct samples based on real user historical travels as a data source, which is equivalent to the user's own travel behavior in actual travel. In other words, since the actual arrival point in the user's historical travel is the arrival location point that the user personally chooses to conveniently enter the POI site, the arrival location point scoring model can be trained based on real user historical travels, and the score of the trained arrival location point scoring model is more accurate. Furthermore, the arrival location point scoring model can be used to explore the arrival location point that is convenient for the user to enter the POI site according to the POI location point.

[0029] The actual arrival location point in this document can refer to a location point actually arrived by a user in a real user trip. It is easy to understand that when a user uses a trip navigation service to make a trip, the trip navigation service will usually plan a navigation path for the user, and the end point of the planned navigation path may not be a suitable arrival location point. The user may not strictly follow the navigation path, but choose a certain location point near the end point of the path as the actual arrival location point according to the actual road conditions. The trip navigation service can record a user's historical trip, and record the POI location point, the actual arrival location point and the planned arrival location point in the historical trip.

[0030] The planned arrival location point is the arrival location point planned by the trip navigation service for the user, which is the end point of the planned navigation path. The purpose of the technical solutions provided by the present disclosure is to improve the ability of the trip navigation service to mine effective arrival location points, so that the arrival location point planned by the trip navigation service for the user is as suitable as possible as the end point of the navigation path, so that the user can smoothly enter the POI site according to the navigation path, and the user does not need to choose other location points near the end point of the navigation path as the actual arrival point according to the actual road conditions, thereby improving the user experience.

[0031] The above technical solutions will be described in detail below in combination with the accompanying drawings.

[0032] Figure 1 An exemplary method for training a POI arrival location point scoring model is provided, including the following steps:

[0033] S100: Obtain a positive sample set and a negative sample set.

[0034] In actual application, the trip navigation service can record user historical trips, and a user historical trip can involve a POI location point that the user wants to go to and an actual arrival location point of the user. In addition, a user historical trip can also involve a planned arrival location point planned by the trip navigation service for the user. The actual arrival location point of the user and the planned arrival location point can be consistent or inconsistent.

[0035] The positive sample usually includes the POI location point and the actual arrival location point in a user historical trip. It is easy to understand that multiple positive samples can be constructed based on different historical trips of different users and based on multiple historical trips of the same user (one user historical trip can generate one positive sample), forming a positive sample set.

[0036] The negative sample usually includes the POI location point and the non-actual arrival location point in a user historical trip. It is easy to understand that multiple negative samples can be constructed based on different historical trips of different users and based on multiple historical trips of the same user (one user historical trip can generate one or more negative samples), forming a negative sample set.

[0037] In some embodiments, a number of historical travel records that meet the positive sample condition can be filtered from the user historical travel record set, and then a corresponding positive sample can be constructed based on each filtered historical travel record.

[0038] In some embodiments, the case where the historical travel record meets the positive sample condition can be that the travel speed corresponding to the end of the trip of the user in the historical travel record is lower than the preset speed. It is easy to understand that the basis for the travel navigation service to determine that the user ends the trip is that the user exits the navigation path or the user reaches the end of the navigation path. However, in actual application, the user exiting the navigation path does not necessarily mean that the user has really ended the trip. One possible case is that the user has not ended the trip and is still traveling, but has exited the navigation path in advance. In this case, the travel navigation service cannot record the actual arrival location point of the user. Therefore, when the travel speed corresponding to the end of the trip of the user is relatively high, it can be determined that the user has not actually ended the trip and is still traveling, and thus the historical travel record recorded by the user cannot be used as a positive sample.

[0039] In addition, the case where the historical travel record meets the positive sample condition can also be that the historical travel record contains the same POI location point and actual arrival location point as at least one other historical travel record. It is easy to understand that if the same user selects the same actual arrival location point for a POI location point more than twice, then the actual arrival location point is likely to be a suitable arrival location point for the end of the navigation track (otherwise, the user is unlikely to select the actual arrival location point at least twice).

[0040] In addition, the number of positive samples can also be expanded. Specifically, for any two historical travel records, if the distance between the POI location points in the two historical travel records is less than a preset distance, a corresponding positive sample can be constructed based on the POI location point in one of the historical travel records and the actual arrival location point in the other historical travel record. It is easy to understand that two POI location points that are relatively close to each other can share an actual arrival location point. In this way, the two POI location points that are relatively close to each other can exchange their arrival location points to generate more positive samples.

[0041] In some embodiments, arrival point sampling can be performed near the POI location point in the filtered historical travel record. Then, in a case where it is determined that the sampled arrival point is not the actual arrival location point in the historical travel record, a corresponding negative sample can be constructed based on the POI location point in the historical travel record and the sampled arrival point. In the driving travel scenario, the above-mentioned arrival point sampling manner can be to randomly sample a parking arrival point on the road near the POI location point.

[0042] S102: training the arrival location point scoring model based on the positive sample set and the negative sample set.

[0043] The idea of training the arrival location point scoring model in the present disclosure is inspired by the idea of training the click model in the search algorithm. The POI location point in a user trip is regarded as a question that the user wants to search, and the arrival location point in the user trip is regarded as an answer searched for the question. The actual arrival point is the answer that meets the user's demand, and the non-actual arrival point is the answer that does not meet the user's demand.

[0044] The algorithm structure of the arrival location point scoring model can be flexibly set. In some embodiments, the algorithm structure of the arrival location point scoring model can be set to include a deep neural network and an activation function. The deep neural network is used to convert sample features into mapping features, and the activation function is used to convert the mapping features into scores of the arrival location point. For example, the algorithm structure of the arrival location point scoring model can be set to a DeepFM structure, which includes a Deep network (a kind of deep neural network) and a sigmoid activation function. The Deep network can usually better extract the relevance information between the POI location point and the arrival point in the sample.

[0045] In performing step S102, usually the sample features (sample features of positive samples or sample features of negative samples) are taken as the model input, and the corresponding model output is the score of the arrival location point (the actual arrival point in the positive sample or the non-actual arrival point in the negative sample) in the sample. The highest score is taken as the label of the positive sample, and the lowest score is taken as the label of the negative sample. Wherein, the highest score can be 1, and the lowest score can be 0. It is easy to understand that the score of the arrival location point can be understood as the probability that the arrival location point is suitable as the end point of the navigation path corresponding to the POI location point. The greater the probability, the more suitable it is.

[0046] The sample features of the sample can be flexibly selected. In some embodiments, the sample features of the positive sample and the negative sample can include: the features of the POI location point in the sample, and the features of the arrival location point in the sample. The arrival location point here can represent the actual arrival location point or the non-actual arrival location point.

[0047] Further, the features of the POI location point include at least one of the following: the administrative division to which the POI location point belongs;

[0048] the functional category of the place corresponding to the POI location point; the operation state of the place corresponding to the POI location point; whether the place corresponding to the POI location point is a bottom shop.

[0049] The features of the arrival location point can include at least one of the following: an administrative level of a road to which the arrival location point belongs; a driving direction of the road to which the arrival location point belongs, the driving direction being one-way or two-way; a width of the road to which the arrival location point belongs; a road type of the road to which the arrival location point belongs, the road type being an intra-organization road or an extra-organization road.

[0050] In addition, the sample features of the positive samples and the negative samples can further include features representing a correlation between the POI location point and the arrival location point in the sample. The correlation features can include at least one of the following: a distance between the POI location point and the arrival location point; whether the POI location point and the arrival location point belong to different roads.

[0051] Further, the positive samples and the negative samples can further include a planned arrival location point provided by the travel navigation service in a historical trip of the user. Accordingly, the features representing the correlation between the POI location point and the arrival location point in the sample can further include a distance between the actual arrival location point or the non-actual arrival location point and the planned arrival location point.

[0052] In actual applications, the features representing the correlation between the POI location point and the arrival location point in the sample can cause a certain deviation to the effect of training the arrival location point scoring model. For example, a user usually selects an actual arrival location point near a planned arrival location point provided by the travel navigation service, but the arrival location point near the planned arrival location point can not be the most suitable one, and the most suitable arrival location point has no necessary relationship with the distance to the planned arrival location point. For another example, an arrival location point near a POI location point can not be the most suitable one, and the most suitable arrival location point also has no necessary relationship with the distance to the POI location point.

[0053] To correct the above bias in training, in one training iteration, the sample features can be input into a reaching location point scoring model to output a score for the reaching location point in the sample; and one or more of the features representing the association between the POI location point and the reaching location point in the sample can be input into an auxiliary scoring model to output a score for the reaching location point in the sample. The score output by the reaching location point scoring model can then be multiplied by the score output by the auxiliary scoring model to obtain a score product. The reaching location point scoring model and the auxiliary scoring model can then be adjusted according to the gap between the sample label and the score product, and the next training iteration can then be entered. For example, the features representing the association between the POI location point and the reaching location point can include the distance between the POI location point and the reaching location point, whether the POI location point and the reaching location point belong to different roads, the distance between the actual reaching location point or the non-actual reaching location point and the planned reaching location point, etc. One or more of these features (e.g., the distance between the actual reaching location point or the non-actual reaching location point and the planned reaching location point) can be input into the auxiliary scoring model.

[0054] In this way, the trained reaching location point scoring model can sufficiently learn the bias caused by the features representing the association between the POI location point and the reaching location point, and has the ability to overcome this bias.

[0055] In addition, considering that the actual reaching point in the user's historical trip that actually occurred can not be the most suitable actual reaching point, after steps S100-S102 are performed, the reaching location point scoring model can be further fine-tuned based on a manually annotated sample set. The manually annotated sample can include a POI location point and a reaching location point that is most suitable as the end point of a navigation path. The score of the fine-tuned reaching location point scoring model can be more accurate.

[0056] The algorithm structure of the auxiliary scoring model described above can be the same as that of the reaching location point scoring model, or can be different (e.g., the algorithm structure of the auxiliary scoring model can be simpler and use a certain linear algorithm structure).

[0057] Figure 2 An exemplary training architecture diagram for training a reaching location point scoring model for a POI is provided. As shown in FIG. 1, the training architecture diagram includes a reaching location point scoring model 100 and an auxiliary scoring model 200. The reaching location point scoring model 100 can be trained to output a score for a reaching location point in a sample based on the features of the sample. The auxiliary scoring model 200 can be trained to output a score for the reaching location point in the sample based on one or more of the features representing the association between the POI location point and the reaching location point in the sample. The score output by the reaching location point scoring model 100 can be multiplied by the score output by the auxiliary scoring model 200 to obtain a score product. The reaching location point scoring model 100 and the auxiliary scoring model 200 can then be adjusted according to the gap between the sample label and the score product, and the next training iteration can then be entered. Figure 2As shown, the training architecture includes a POI location point scoring model and an auxiliary scoring model. The input of the POI location point scoring model is the feature of the POI location point in the sample, the feature of the arrival location point in the sample, and the relevance feature between the POI location point and the arrival location point in the sample. The input of the auxiliary scoring model is one or more of the relevance features between the POI location point and the arrival location point. The score output by the arrival location point scoring model is multiplied by the score output by the auxiliary scoring model. The score product can be understood as a score after bias correction. The closer the score product is to the score in the sample label, the better the training effect of the model.

[0058] Figure 3 An exemplary method for determining a navigation path is provided, including:

[0059] S300: Obtain a POI location point specified by a user, and obtain a plurality of candidate arrival location points corresponding to the POI location point.

[0060] The plurality of candidate arrival location points corresponding to the POI location point can be a plurality of arrival location points planned by a travel navigation service based on a certain planning strategy. The most suitable candidate arrival location point can be discovered from the candidate arrival location points by using the discovery capability of the model.

[0061] S302: For each candidate arrival location point, call an arrival location point scoring model according to the POI location point and the candidate arrival location point, and score the candidate arrival location point.

[0062] S304: Determine a navigation path corresponding to the POI location point by taking the candidate arrival location point with the highest score as the end point of the navigation path.

[0063] Of course, the first few candidate arrival points with relatively high scores can also be discovered to form several navigation paths for the user to select.

[0064] In addition, in actual application, the arrival location point scoring model can be quickly iteratively optimized based on the collected user travel data on a regular or irregular basis, so that changes in the entrance of the POI site in the real world can be responded to in a timely manner.

[0065] The present disclosure also provides a computer readable storage medium, such as a computer readable storage medium 140. Figure 4 As shown, the computer program is stored on the medium 140, and the program is executed by a processor to implement the method of the embodiments of the present disclosure.

[0066] The present disclosure also provides a computing device including a memory and a processor. The memory is configured to store computer instructions executable on the processor, and the processor is configured to implement the method of the embodiments of the present disclosure when executing the computer instructions.

[0067] Figure 5 is a structural schematic diagram of a computing device provided by the present disclosure, which can include but is not limited to: a processor 151, a memory 152, a bus 153 connecting different system components including the memory 152 and the processor 151.

[0068] The memory 152 stores computer instructions that can be executed by the processor 151, so that the processor 151 can execute the method of any embodiment of the present disclosure. The memory 152 can include a random access memory unit RAM 1521, a cache memory unit 1522 and / or a read-only memory unit ROM 1523. The memory 152 can also include a program tool 1525 having a set of program modules 1524, which include but are not limited to: an operating system, one or more application programs, other program modules and program data, which can include one or more combinations of network environment implementations.

[0069] The bus 153 can include, for example, a data bus, an address bus and a control bus, etc. The computing device 15 can also communicate with an external device 155, which can be, for example, a keyboard, a Bluetooth device, etc., through an I / O interface 154. The computing device 150 can also communicate with one or more networks, for example, a local area network, a wide area network, a public network, etc., through a network adapter 156. As shown, the network adapter 156 can also communicate with other modules of the computing device 15 through the bus 153.

[0070] In addition, although the operations of the method of the present disclosure are described in a particular order in the accompanying drawings, this does not require or imply that the operations must be performed in this particular order, or that all of the shown operations must be performed to achieve the desired result. Additionally or alternatively, certain steps can be omitted, a plurality of steps can be combined into one step, and / or one step can be divided into a plurality of steps.

[0071] Although the spirit and principles of the present disclosure have been described with reference to several specific embodiments, it should be understood that the present disclosure is not limited to the disclosed specific embodiments, and the division of aspects does not mean that the features in these aspects cannot be combined to benefit. This division is only for the convenience of expression. The present disclosure is intended to cover various modifications and equivalent arrangements included in the spirit and scope of the appended claims.

[0072] The systems, apparatuses, modules, or units disclosed in the above embodiments can be specifically implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, the computer can be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0073] For the sake of description, the above apparatuses are described in various units by functions respectively. Of course, the functions of the units can be implemented in one or more software and / or hardware in implementing the present specification.

[0074] Those skilled in the art should understand that embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROMs, optical storage devices, etc.) containing computer-usable program code.

[0075] The present application is described with reference to flowcharts and / or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, as well as combinations of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing apparatus to produce a machine, so that the instructions, which are executed via the processor of the computer or other programmable data processing apparatus, generate a means for implementing the functions specified in the flowcharts and / or block diagrams. Figure 1 The functions specified in a flow or multiple flows and / or blocks Figure 1 The apparatus that implements the functions specified in a flow or multiple flows and / or blocks.

[0076] The present specification can be described in the general context of computer-executable instructions, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, and the like that perform particular tasks or implement particular abstract data types. The present specification can also be practiced in distributed computing environments where tasks are performed by remote processing devices that are connected through a communication network. In a distributed computing environment, program modules can be located in both local and remote computer storage media including storage devices.

[0077] These computer program instructions can also be stored in a computer- readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the Figure 1 function specified in the flow or flows and / or blocks Figure 1 of the block or blocks.

[0078] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the Figure 1 function specified in the flow or flows and / or blocks Figure 1 of the block or blocks. In one typical configuration, the computer includes one or more processors (CPUs), memory, input / output interfaces, and network interfaces.

[0079] The memory can include non-persistent memory and / or volatile memory, such as random access memory (RAM) about which the computer stores the information. The memory is an example of computer readable media.

[0080] Computer readable media includes permanent and non-permanent, moveable and non- moveable media which can be implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile discs (DVDs) or other optical storage, magnetic cassettes, magnetic disks storage, quantum memory, graphene-based storage media, or other magnetic storage devices, or any other medium which can be used to store information which can be accessed by a computing device. According to the definition herein, computer readable media does not include transitory media, such as modulated data signals and carrier waves.

[0081] It is also to be noted that the terms "comprising", "including", and "having" or variations thereof herein, are intended to be open-ended terms that specify the presence of the stated elements but do not preclude the presence of additional elements. It is also to be noted that the term "if' as used herein, encompasses the meanings of both "if' and "when," and that the term "including" as used herein, means "including, but not limited to."

[0082] The above-described embodiments of the present specification are described in connection with the respective features. Other embodiments within the scope of the following claims can be apparent to those of ordinary skill in the art from the description herein. In some cases, the actions recited in the claims can be performed in a different order and still achieve desirable results. In some cases, steps or functions can be combined into fewer steps or functions without losing the desired effect. Other embodiments can be implemented using different materials, or manufacturing techniques, without deviating from the scope of the claims. In some cases, additional steps can be added before, after, or in between the steps described herein without deviating from the scope of the claims.

[0083] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of embodiments of the present specification. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "comprises" and / or "comprising," when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0084] It will be understood that, although the terms first, second, third, etc. can be used herein to describe various information, these terms are used only to distinguish one from another information. For example, a first information could be termed a second information, and, similarly, a second information could be termed a first information, without departing from the scope of embodiments of the present specification. As used herein, the term "if' can be construed to mean "when" or "if," depending on the context. It will be further understood that the terms "comprises" and / or "comprising," when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0085] Various embodiments are described herein with reference to the following items, which are not necessarily mutually exclusive. The embodiments described with reference to one item can be combined with the embodiments described with reference to another item to describe additional embodiments. The following items are provided as separate embodiments for clarity. However, it should be understood that the embodiments described with reference to one item can be combined with the embodiments described with reference to another item to describe additional embodiments. The following items are provided as separate embodiments for clarity. However, it should be understood that the embodiments described with reference to one item can be combined with the embodiments described with reference to another item to describe additional embodiments.

[0086] The above description is merely illustrative of the embodiments of the present disclosure and is not in any way intended to limit the present disclosure. It should be understood by one of ordinary skill in the art that any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present disclosure should be included in the scope of the present disclosure.

Claims

1. A method for training a POI arrival location point scoring model, comprising: obtaining a positive sample set and a negative sample set; wherein the positive sample comprises a POI location point and an actual arrival location point in a user historical trip, and the positive sample is constructed based on a POI location point in one historical trip record and an actual arrival location point in another historical trip record if the distance between the two POI location points is less than a preset distance; the negative sample comprises a POI location point and a non-actual arrival location point in a user historical trip; training the arrival location point scoring model based on the positive sample set and the negative sample set; wherein the sample feature is used as the model input, and the corresponding model output is the score of the arrival location point in the sample. 2.The method of claim 1, wherein obtaining the positive sample set comprises: selecting a plurality of historical trip records that meet the positive sample condition from the user historical trip record set; constructing a corresponding positive sample based on each selected historical trip record; wherein the historical trip record meets the positive sample condition in the following cases: the corresponding travel speed when the user ends the trip in the historical trip record is lower than a preset speed; and / or the historical trip record and at least one other historical trip record contain the same POI location point and actual arrival location point. 3.The method of claim 2, wherein obtaining the negative sample set comprises: sampling arrival points near the POI location point in the selected historical trip record; constructing a corresponding negative sample based on the POI location point in the historical trip record and the sampled arrival point if it is determined that the sampled arrival point is not the actual arrival location point in the historical trip record.

4. The method of claim 1, wherein, The sample features of the positive sample and the negative sample include the features of the POI location point in the sample and the features of the arrival location point in the sample; the arrival location point includes the actual arrival location point or the non-actual arrival location point.

5. The method of claim 4, wherein, The features of the POI location point include at least one of the following: the administrative division to which the POI location point belongs; the functional category of the place corresponding to the POI location point; the operation status of the place corresponding to the POI location point; whether the place corresponding to the POI location point is a bottom shop.

6. The method of claim 4, wherein, The features of the arrival location point include at least one of the following: the administrative level of the road to which the arrival location point belongs; the driving direction of the road to which the arrival location point belongs; the driving direction is one-way or two-way; the width of the road to which the arrival location point belongs; the road type of the road to which the arrival location point belongs; the road type is an internal road or an external road.

7. The method of claim 4, wherein, The sample features of the positive sample and the negative sample further include a feature representing the association between the POI location point and the arrival location point in the sample.

8. The method of claim 7, wherein, The feature representing the association between the POI location point and the arrival location point in the sample includes at least one of the following: the distance between the POI location point and the arrival location point; whether the POI location point and the arrival location point belong to different roads.

9. The method of claim 7, wherein, The positive sample and the negative sample further include a planned arrival location point provided by a travel navigation service in a user historical trip. The features characterizing the association between the POI location point and the arrival location point in the sample also include: a distance between the actual or non-actual arrival location point and the planned arrival location point.

10. The method of claim 1, wherein, The algorithm structure of the arrival location point scoring model includes a deep neural network and an activation function; the deep neural network is used to convert the sample features into mapping features, and the activation function is used to convert the mapping features into a score for the arrival location point.

11. The method of claim 7, wherein the arrival location point scoring model is trained based on the positive sample set and the negative sample set, comprising: in one training iteration, inputting the sample features into the arrival location point scoring model to output a score for the arrival location point in the sample, and inputting one or more of the features characterizing the association between the POI location point and the arrival location point in the sample into the auxiliary scoring model to output a score for the arrival location point in the sample; multiplying the score output by the arrival location point scoring model and the score output by the auxiliary scoring model to obtain a score product; adjusting the arrival location point scoring model and the auxiliary scoring model according to the gap between the sample label and the score product, and then entering the next training iteration.

12. A method for determining a navigation path, applied to a travel navigation service, the method comprising: obtaining a POI location point specified by a user, and obtaining a plurality of candidate arrival location points corresponding to the POI location point; for each candidate arrival location point, calling the arrival location point scoring model trained by the method of any one of claims 1-11 to score the candidate arrival location point based on the POI location point and the candidate arrival location point; determining a navigation path corresponding to the POI location point by taking the candidate arrival location point with the highest score as the end point of the navigation path.

13. A computing device comprising a memory and a processor; the memory is used to store computer instructions executable on the processor, and the processor is used to implement the method of any one of claims 1-12 when executing the computer instructions.

Images