Navigation route preference acquisition method and device, navigation route recommendation method and device

Abstract

The disclosed embodiments disclose a method and device for obtaining navigation route preferences, and a method and device for recommending navigation routes. The method for obtaining navigation route preferences includes: obtaining the navigation route with the highest coverage rate among the navigation routes of the same navigation object at different times as the target navigation route; obtaining the route characteristics of the target navigation route and the statistical environmental characteristics associated with the route characteristics as the route preference characteristics of the target navigation route at the corresponding time; obtaining a route preference feature sequence based on the route preference features corresponding to different times; using the route preference feature sequence as the input of a machine learning model, and obtaining the route preference vector of the navigation object through the machine learning model. A new route feature representation method can be used to complete route preference learning, thereby truly and comprehensively reflecting the personalized route preferences of the navigation object.

Description

Technical Field

[0001] The present disclosure relates to the field of navigation technology, and in particular to a method and device for obtaining navigation route preferences, and a method and device for recommending navigation routes. Background Art

[0002] Currently, applications with map navigation functions, such as online ride-hailing applications, are becoming increasingly important for users' travel. When using the navigation function, different users have different preferences for navigation routes, even for the same starting and ending points. For example, some users are more concerned about the time cost of travel, so they expect the navigation routes provided by the application to include longer highways. For another example, some users are more sensitive to costs, so they prefer the navigation routes provided by the application to include roads with low or no tolls. For another example, some users dislike bumpy roads and expect the roads included in the navigation routes provided by the application to have better road maintenance.

[0003] Existing technologies allow users to manually set route recommendation criteria within apps, such as prioritizing highways, minimizing tolls, and prioritizing major roads. However, these settings are limited and fail to fully reflect users' actual route needs. Consequently, the recommended navigation routes often fail to meet the personalized needs of the users being guided. To address this issue, a new technical solution is needed that can more comprehensively and accurately describe users' personalized preferences for navigation routes. Summary of the Invention

[0004] In order to solve the problems in the related art, the embodiments of the present disclosure provide a navigation route preference acquisition method and device, a navigation route recommendation method and device, and a computer program product, which can truly and comprehensively reflect the user's route preferences, and can recommend routes based on the real personalized needs of the user as the navigation object, or determine whether the navigation-planned route meets the real personalized needs of the user as the navigation object.

[0005] In a first aspect, an embodiment of the present disclosure provides a method for obtaining navigation route preferences, wherein the method includes:

[0006] Obtain the navigation route with the largest coverage rate among the navigation routes of the same navigation object at different times as the target navigation route;

[0007] Acquiring route characteristics of the target navigation route and statistical environmental characteristics associated with the route characteristics as route preference characteristics at a corresponding moment of the target navigation route;

[0008] obtaining a route preference feature sequence based on the route preference features corresponding to different moments;

[0009] The route preference feature sequence is used as an input of a machine learning model, and a route preference vector of the navigation object is obtained through the machine learning model.

[0010] In conjunction with the first aspect, in a first implementation of the first aspect of the present disclosure, taking the route preference feature sequence as input to a machine learning model and obtaining the route preference vector of the navigation object through the machine learning model includes:

[0011] Encoding the route preference feature sequences at different moments using a self-attention model to obtain weights of the route preference features at different moments;

[0012] The route preference features at different moments are weighted based on the weights, and the weighted results are aggregated to obtain a route preference vector of the navigation object representing the route preference of the navigation object.

[0013] In combination with the first implementation manner of the first aspect, in a second implementation manner of the first aspect of the present disclosure, the weights of the route preference features at different moments represent the similarities between the route preference features at different moments.

[0014] In combination with the first aspect, and any one of the first implementation to the second implementation of the first aspect, in a third implementation of the first aspect of the present disclosure, the method further includes:

[0015] Obtaining the navigation route with the largest coverage rate among the navigation routes of the same navigation object at the target time as the target navigation route;

[0016] Acquiring route characteristics of the target navigation route and statistical environmental characteristics associated with the route characteristics as route preference characteristics at the target moment;

[0017] performing a route preference prediction of the navigation object based on the route preference vector of the navigation object output by the machine learning model and the route preference feature at the target time, to obtain a route preference prediction result of the navigation object;

[0018] Adjust the parameters of the machine learning model based on the route preference prediction result of the navigation object.

[0019] In combination with the first aspect, and any one of the first implementation manner to the second implementation manner of the first aspect of the present disclosure, in a fourth implementation manner of the first aspect, the route feature is a route feature vector including multidimensional route features, the statistical environment feature is a statistical environment feature vector including multidimensional statistical environment features, and the route preference feature includes the route feature vector and the statistical environment feature vector.

[0020] The statistical environment feature vector includes the maximum feature value and the minimum feature value of each dimensional feature in the route feature vector among the route features of the corresponding dimensions of all planned routes.

[0021] In a second aspect, an embodiment of the present disclosure provides a navigation route preference acquisition device, wherein the device includes:

[0022] A target navigation route acquisition module is configured to acquire the navigation route with the largest coverage rate among the navigation routes of the same navigation object at different times as the target navigation route;

[0023] a route preference feature acquisition module configured to acquire the route features of the target navigation route and the statistical environment features associated with the route features as the route preference features of the target navigation route at a corresponding moment;

[0024] a route preference feature sequence acquisition module, configured to acquire a route preference feature sequence based on the route preference features corresponding to different moments;

[0025] The route preference vector acquisition module is configured to use the route preference feature sequence as an input of a machine learning model, and obtain the route preference vector of the navigation object through the machine learning model.

[0026] In conjunction with the second aspect, in a first implementation of the second aspect of the present disclosure, the route preference vector acquisition module is further configured to:

[0027] Encoding the route preference feature sequences at different moments using a self-attention model to obtain weights of the route preference features at different moments;

[0028] The route preference features at different moments are weighted based on the weights, and the weighted results are aggregated to obtain a route preference vector of the navigation object representing the route preference of the navigation object.

[0029] In a third aspect, an embodiment of the present disclosure provides a navigation route recommendation method, wherein the method includes:

[0030] Get the navigation route to be recommended to the navigation object;

[0031] determining a route preference vector of the navigation object;

[0032] Based on the route preference vector of the navigation object, a recommended order of the navigation routes is determined, wherein the route preference vector of the navigation object is determined by any one of the first aspect and the first to fourth implementations of the first aspect.

[0033] In a fourth aspect, an embodiment of the present disclosure provides a navigation route recommendation device, wherein the device includes:

[0034] A navigation route acquisition module is configured to acquire a navigation route to be recommended to a navigation object;

[0035] The navigation route preference acquisition device of the second aspect or the first implementation of the second aspect is configured to determine a route preference vector of the navigation object;

[0036] A recommended order determination module is configured to determine a recommended order of the navigation routes based on the route preference vector of the navigation object, wherein the route preference vector of the navigation object is determined by the navigation route preference acquisition device of the second aspect or the first implementation manner of the second aspect.

[0037] In a fifth aspect, a computer program product is provided in an embodiment of the present disclosure, comprising computer instructions, which, when executed by a processor, implement the method as described in the first aspect, the first to fourth implementations of the first aspect, and the third aspect.

[0038] The technical solutions provided by the embodiments of the present disclosure may have the following beneficial effects:

[0039] According to the technical solution provided by the embodiment of the present disclosure, the navigation route with the largest coverage rate among the navigation routes of the same navigation object at different times is obtained as the target navigation route; the route characteristics of the target navigation route and the statistical environmental characteristics associated with the route characteristics are obtained as the route preference characteristics of the target navigation route at the corresponding time; based on the route preference characteristics corresponding to different times, a route preference feature sequence is obtained; the route preference feature sequence is used as the input of a machine learning model, and the route preference vector of the navigation object is obtained through the machine learning model. A new route feature representation method can be used to complete route preference learning, thereby truly and comprehensively reflecting the personalized route preference of the navigation object. Moreover, it is helpful to make route recommendations based on the real personalized needs of the user as the navigation object, or to help determine whether the route planned by the navigation meets the real personalized needs of the user as the navigation object.

[0040] According to the technical solution provided by the embodiment of the present disclosure, the route preference feature sequence is used as the input of the machine learning model, and the route preference vector of the navigation object is obtained through the machine learning model, including: using the self-attention model to encode the route preference feature sequences at different times to obtain weights of the route preference features at different times; weighting the route preference features at different times based on the weights, and aggregating the weighted results to obtain the route preference vector of the navigation object representing the route preference of the navigation object. The self-attention mechanism can be used to reduce the noise of the historical navigation sequence of the user as the navigation object, and solves the problem of how to describe the user's personal preference for routes in the personalized route recommendation scenario.

[0041] According to the technical solution provided by the embodiments of the present disclosure, the similarity between the route preference features at different times is represented by the weights of the route preference features at different times. The self-attention mechanism can be used to reduce noise in the historical navigation sequence of the user as the navigation target. This also solves the problem of how to describe the user's personal preferences for routes in the personalized route recommendation scenario.

[0042] According to the technical solution provided by the embodiment of the present disclosure, the method further includes: obtaining the navigation route with the largest coverage rate among the navigation routes of the same navigation object at the target moment as the target navigation route; obtaining the route characteristics of the target navigation route and the statistical environmental characteristics associated with the route characteristics as the route preference characteristics at the target moment; predicting the route preference of the navigation object based on the route preference vector of the navigation object output by the machine learning model and the route preference characteristics of the target moment to obtain the route preference prediction result of the navigation object; adjusting the parameters of the machine learning model based on the route preference prediction result of the navigation object, and adopting a new route feature representation method to complete route preference learning, thereby truly and comprehensively reflecting the personalized route preference of the navigation object. Moreover, it is helpful to make route recommendations based on the real personalized needs of the user as the navigation object, or to help determine whether the route planned by the navigation meets the real personalized needs of the user as the navigation object.

[0043] According to the technical solution provided by the embodiment of the present disclosure, the route feature is a route feature vector including multi-dimensional route features, the statistical environment feature is a statistical environment feature vector including multi-dimensional statistical environment features, and the route preference feature includes the route feature vector and the statistical environment feature vector, wherein the statistical environment feature vector includes the maximum eigenvalue and the minimum eigenvalue of the route features of the corresponding dimensions of all planned routes in the route feature vector for each dimensional feature, a new route feature representation method can be used to complete route preference learning, thereby truly and comprehensively reflecting the personalized route preference of the navigation object.

[0044] According to the technical solution provided by the embodiment of the present disclosure, a target navigation route acquisition module is configured to acquire the navigation route with the largest coverage rate among the navigation routes of the same navigation object at different times as the target navigation route; a route preference feature acquisition module is configured to acquire the route features of the target navigation route and the statistical environmental features associated with the route features as the route preference features of the target navigation route at the corresponding moment; a route preference feature sequence acquisition module is configured to acquire a route preference feature sequence based on the route preference features corresponding to different moments; a route preference vector acquisition module is configured to use the route preference feature sequence as the input of a machine learning model, and obtain the route preference vector of the navigation object through the machine learning model. A new route feature representation method can be used to complete route preference learning, thereby truly and comprehensively reflecting the personalized route preferences of the navigation object. Moreover, it is helpful to make route recommendations based on the real personalized needs of the user as the navigation object, or to help determine whether the route planned by the navigation meets the real personalized needs of the user as the navigation object.

[0045] According to the technical solution provided by the embodiment of the present disclosure, the route preference vector acquisition module is further configured to: encode the route preference feature sequences at different times using a self-attention model to obtain weights of the route preference features at different times; weight the route preference features at different times based on the weights, and aggregate the weighted results to obtain the route preference vector of the navigation object representing the route preference of the navigation object. The self-attention mechanism can be used to reduce noise on the historical navigation sequence of the user who is the navigation object, and the problem of how to describe the user's personal preference for routes in the personalized route recommendation scenario is solved.

[0046] According to the technical solution provided by the embodiment of the present disclosure, a navigation route to be recommended to a navigation object is obtained; a route preference vector of the navigation object is determined; and a recommendation order of the navigation route is determined based on the route preference vector of the navigation object. The route preference vector of the navigation object is determined by the navigation route preference acquisition method. A new route feature representation method can be used to complete route preference learning, thereby truly and comprehensively reflecting the personalized route preferences of the navigation object. Furthermore, route recommendations can be made based on the actual personalized needs of the user who is the navigation object, or it can be determined whether the route planned by the navigation object meets the actual personalized needs of the user who is the navigation object. A new route feature representation method can be used to complete route preference learning, thereby truly and comprehensively reflecting the personalized route preferences of the navigation object. Furthermore, route recommendations can be made based on the actual personalized needs of the user who is the navigation object.

[0047] According to the technical solution provided by the embodiments of the present disclosure, a navigation route acquisition module is configured to acquire a navigation route to be recommended to a navigation subject; a navigation route preference acquisition device is configured to determine a route preference vector for the navigation subject; and a recommendation order determination module is configured to determine a recommendation order for the navigation routes based on the route preference vector for the navigation subject. The navigation route preference acquisition device of the second aspect or the first implementation of the second aspect determines the route preference vector for the navigation subject, and a new route feature representation method can be used to complete route preference learning, thereby truly and comprehensively reflecting the personalized route preferences of the navigation subject. Furthermore, route recommendations can be made based on the actual personalized needs of the user serving as the navigation subject.

[0048] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Other features, objectives and advantages of the present disclosure will become more apparent through the following detailed description of non-limiting embodiments in conjunction with the accompanying drawings. In the accompanying drawings:

[0050] Figure 1 A flowchart illustrating a method for obtaining navigation route preferences according to an embodiment of the present disclosure is shown;

[0051] Figure 2 An exemplary schematic diagram showing route preference characteristics of a user at different times obtained during implementation of a method for obtaining navigation route preferences according to an embodiment of the present disclosure;

[0052] Figure 3 A schematic diagram illustrating a self-attention model training method of an algorithm of a navigation route preference acquisition method according to an embodiment of the present disclosure;

[0053] Figure 4 A structural block diagram of a navigation route preference acquisition device according to an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0054] Hereinafter, exemplary embodiments of the present disclosure will be described in detail with reference to the accompanying drawings so that those skilled in the art can easily implement them. In addition, for the sake of clarity, parts not related to the description of the exemplary embodiments are omitted in the accompanying drawings.

[0055] In the present disclosure, it should be understood that terms such as "including" or "having" are intended to indicate the presence of labels, numbers, steps, actions, components, parts, or combinations thereof disclosed in the present specification, and are not intended to exclude the possibility that one or more other labels, numbers, steps, actions, components, parts, or combinations thereof exist or are added.

[0056] It should also be noted that, in the absence of conflict, the embodiments and labels in the embodiments of the present disclosure can be combined with each other. The present disclosure will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0057] To facilitate understanding of the embodiments of the present disclosure, some concepts mentioned in the embodiments of the present disclosure are explained below.

[0058] User's actual route: The sequence of recorded tracks from the starting point to the end point of a user's use of an application with a map navigation function (for example, a GPS track recorded by an application with a map navigation function) can be called the user's actual route.

[0059] (Navigation) Route Planning: When a user uses an app with map navigation functionality, the app will provide several routes from a starting point to a destination for the user to choose from. The process of calculating a route by the app is called navigation planning, and the calculated route is called a planning result. A planning result can include one or more (navigation) planned routes, also referred to as navigation routes.

[0060] Actual Travel Coverage: This is calculated by dividing the length of the overlap between the navigation route and the user's actual route by the length of the user's actual route. As a rule of thumb, a higher actual travel coverage indicates a more comprehensive coverage of the navigation route for the user's needs.

[0061] Historical navigation sequence: When a user uses an app with map navigation functionality, the entire process from clicking to plan a route to closing the navigation process after reaching the destination is called a navigation session. For users, repeated use of navigation tools generates a series of planned routes and actual routes. These planned routes and actual routes are linked in chronological order to form the user's historical navigation sequence.

[0062] Self-Attention: An open-source technology that can be used for sequence encoding tasks in machine learning. The self-attention mechanism is a variant of the attention mechanism that reduces reliance on external information and is better at capturing internal correlations in data or features.

[0063] Embedding: An open source technology that can transform hard-coded data into vector representations in a low-dimensional space.

[0064] Related art applications with map navigation functionality struggle to recommend routes based on the actual personalized needs of the users being navigated. While recommendations based on user preferences exist in other related art fields, such approaches are not suitable for personalized route recommendations in applications with map navigation functionality. For example, in traditional recommendation domains such as e-commerce, advertising, and video websites, describing a user's personalized preferences based on their historical behavior sequences is a common and universal approach. For example, in e-commerce, a user's historical purchases often represent their shopping preferences. Consumers who frequently purchase clothing and those who frequently purchase electronics have distinctly different shopping preferences. In these domains, representation of products, advertisements, and videos is often achieved through pre-training and embedding techniques (i.e., each product has a unique ID, and embedding is then used to generate its feature representation). However, in the field of route recommendation, embedding cannot be used for representation due to the infinite number of routes.

[0065] Therefore, in the embodiment of the present disclosure, when each navigation is directly used, the planned route with the largest coverage is used to represent the user's route preference during this navigation.

[0066] According to the technical solution provided by the embodiment of the present disclosure, the navigation route with the largest coverage rate among the navigation routes of the same navigation object at different times is obtained as the target navigation route; the route characteristics of the target navigation route and the statistical environmental characteristics associated with the route characteristics are obtained as the route preference characteristics of the target navigation route at the corresponding time; based on the route preference characteristics corresponding to different times, a route preference feature sequence is obtained; the route preference feature sequence is used as the input of a machine learning model, and the route preference vector of the navigation object is obtained through the machine learning model. A new route feature representation method can be used to complete route preference learning, thereby truly and comprehensively reflecting the personalized route preference of the navigation object. Moreover, it is helpful to make route recommendations based on the real personalized needs of the user as the navigation object, or to help determine whether the route planned by the navigation meets the real personalized needs of the user as the navigation object.

[0067] Figure 1 FIG. 1 is a flow chart showing a method for obtaining navigation route preferences according to an embodiment of the present disclosure. Figure 1 As shown, the navigation route preference acquisition method includes steps S101, S102, S103, and S104.

[0068] In step S101, a navigation route with the largest coverage rate among navigation routes of the same navigation object at different times is obtained as a target navigation route.

[0069] In step S102, the route characteristics of the target navigation route and the statistical environment characteristics associated with the route characteristics are obtained as the route preference characteristics of the target navigation route at a corresponding moment.

[0070] In step S103, a route preference feature sequence is acquired based on the route preference features corresponding to different moments.

[0071] In step S104, the route preference feature sequence is used as an input of a machine learning model, and a route preference vector of the navigation object is obtained through the machine learning model.

[0072] In one embodiment of the present disclosure, the route characteristics of the navigation route with the highest coverage of the user's actual route and the statistical environmental characteristics associated with the route characteristics can be obtained according to methods known in the relevant art. In one embodiment of the present disclosure, for each user, the historical navigation sequence of the user as the navigation target can be concatenated in order from morning to night.

[0073] The following reference Figure 2 The route preference features of a user (navigation target) at different moments obtained during the implementation of a navigation route preference acquisition method according to an embodiment of the present disclosure are described. Figure 2 An exemplary schematic diagram illustrating route preference characteristics of a user at different times obtained during implementation of a method for obtaining navigation route preferences according to an embodiment of the present disclosure is shown.

[0074] like Figure 2As shown, on the time axis T, for user 210, the navigation route and the route 211 taken by the user are obtained in order from morning to night, thereby obtaining the historical navigation sequence of user 210. On the time axis T, for user 220, the navigation route and the route 221 taken by the user are obtained in order from morning to night, thereby obtaining the historical navigation sequence of user 220. In the historical navigation sequence of user 210 or 220, there may be navigation routes and the route 211 or 221 taken by the user in the order of time t1, t2, t3...tn. For each individual time point, such as t1, the navigation route with the highest coverage rate taken by user 210 or 220 in the planning results during this navigation is selected as the route that best represents the navigation preference of user 210 or 220. The characteristics of the preferred route of this navigation, such as ETA (Estimated Time of Arrival), starting point and end point distance, traffic lights, etc., together with the statistical environmental characteristics associated with the route characteristics of the preferred route of this navigation, constitute the preference characteristics of the user 210 or 220 in this navigation.

[0075] In one embodiment of the present disclosure, the statistical environment characteristics associated with the route characteristics of the preferred navigation route may refer to the maximum and minimum characteristic values ​​among the characteristics corresponding to the characteristics of the preferred navigation route of the multiple navigation routes issued by the application with map navigation function through the planning results. For example, the ETA of this navigation is 1 hour, and the ETAs corresponding to the characteristics of the preferred navigation route of the multiple navigation routes issued by the application with map navigation function through the planning results are 55 minutes, 56 minutes...1 hour and 15 minutes, and 1 hour and 25 minutes, respectively. The maximum characteristic value among the characteristics corresponding to the characteristics of the preferred navigation route of the multiple navigation routes issued by the application with map navigation function through the planning results is 1 hour and 25 minutes, and the minimum characteristic value is 55 minutes. In an embodiment of the present disclosure, the navigation routes other than the preferred route (the navigation route with the highest coverage rate of the user's actual route) in the planning results issued during the current navigation constitute the statistical environment. Therefore, at each moment, the route characteristics of the preferred navigation route and the statistical environment characteristics associated with the route characteristics are together used as the route preference characteristics at that moment.

[0076] In one embodiment of the present disclosure, the reason why statistical environmental features associated with the route features of the navigation preference route are needed when processing the route preference of the navigation object (user) is because when learning the user's route preference, it is necessary to learn the position of each route feature of the preferred route among the corresponding features in the entire statistical environment among multiple navigation routes. In other words, through the statistical environmental features associated with the route features of the navigation preference route, the machine learning model of the embodiment of the present disclosure can learn what kind of features are more likely to be the route features of the user's preferred route. In the embodiment of the present disclosure, the statistical environmental features associated with the route features of the navigation preference route can be considered as a benchmark in a sense, used to measure the route features of the navigation preference route.

[0077] In one embodiment of the present disclosure, a sequence of route preference features acquired in the order of time t1, t2, t3, ..., tn may be used as a route preference feature sequence. Therefore, the route preference feature sequence includes multiple route preference features connected in series in time order.

[0078] In one embodiment of the present disclosure, the route feature is a route feature vector including multidimensional route features, the statistical environment feature is a statistical environment feature vector including multidimensional statistical environment features, and the route preference feature includes the route feature vector and the statistical environment feature vector, wherein the statistical environment feature vector includes the maximum eigenvalue and the minimum eigenvalue of each dimensional feature in the route feature vector among the route features of the corresponding dimensions of all navigation routes.

[0079] In one embodiment of the present disclosure, the route features are represented in vector form, which is convenient for processing by an algorithm. In one embodiment of the present disclosure, since the amount of statistical environmental features associated with the route features of the user's preferred route may be large, if the statistical environmental features of all navigation routes are taken as part of the route preference features, then since a planning result may have many navigation routes, the amount of data of the route preference features with the statistical environmental features of all navigation routes added may be too large to be processed. Therefore, in order to obtain a route preference feature with reasonable dimensions and facilitate data processing, the maximum eigenvalue and minimum eigenvalue of each dimensional feature in the route feature vector among the route features of the corresponding dimensions of all navigation routes are taken as part of the route preference features. Taking the route preference feature including ETA as an example, in addition to including the ETA of the navigation route with the largest route coverage (for example, the aforementioned 1 hour), the route preference feature may also include the maximum eigenvalue of 1 hour and 25 minutes and the minimum eigenvalue of 55 minutes of the ETA of the user's preferred route among the route features of the corresponding dimensions of all navigation routes. In one example of the present disclosure, when the route feature is a route feature vector including 100-dimensional route features, the statistical environment feature vector includes 100 dimensions of maximum eigenvalues ​​and 100 dimensions of minimum eigenvalues ​​in the route features of corresponding dimensions, that is, a route preference feature vector includes 300 dimensions of eigenvalues.

[0080] According to the technical solution provided by the embodiment of the present disclosure, the route feature is a route feature vector including multi-dimensional route features, the statistical environment feature is a statistical environment feature vector including multi-dimensional statistical environment features, and the route preference feature includes the route feature vector and the statistical environment feature vector, wherein the statistical environment feature vector includes the maximum eigenvalue and the minimum eigenvalue of the route features of the corresponding dimensions of all navigation routes in the route feature vector for each dimensional feature, a new route feature representation method can be used to complete route preference learning, thereby truly and comprehensively reflecting the personalized route preference of the navigation object.

[0081] In one embodiment of the present disclosure, step S104 includes: encoding the route preference feature sequences at different times using a self-attention model to obtain weights of the route preference features at different times; weighting the route preference features at different times based on the weights, and aggregating the weighted results to obtain a route preference vector of the navigation object representing the route preference of the navigation object.

[0082] In one embodiment of the present disclosure, the self-attention model can notice the correlation between different features in the route preference feature sequence as the entire input. Therefore, after obtaining the route preference features of the user at each moment and forming a route preference feature sequence, it can be encoded using self-attention technology. The self-attention technology can learn the weight of the route preference features at each time (each moment) in the historical sequence. By aggregating the weighted route preference features, the aggregation result (the route preference vector of the navigation object representing the route preference of the navigation object) can be added to the target prediction task for the most recent navigation (this task can be selected arbitrarily). By training in the target prediction task, a fully learned re-learning module can be obtained.

[0083] According to the technical solution provided by the embodiment of the present disclosure, the route preference feature sequence is used as the input of the machine learning model, and the route preference vector of the navigation object is obtained through the machine learning model, including: using the self-attention model to encode the route preference feature sequences at different times to obtain weights of the route preference features at different times; weighting the route preference features at different times based on the weights, and aggregating the weighted results to obtain the route preference vector of the navigation object representing the route preference of the navigation object. The self-attention mechanism can be used to reduce the noise of the historical navigation sequence of the user as the navigation object, and solves the problem of how to describe the user's personal preference for routes in the personalized route recommendation scenario.

[0084] The following reference Figure 3 The present invention describes a self-attention model training method of an algorithm for a navigation route preference acquisition method according to an embodiment of the present invention. Figure 3 A schematic diagram illustrating a self-attention model training method of an algorithm of a navigation route preference acquisition method according to an embodiment of the present disclosure.

[0085] like Figure 3As shown, the route preference feature sequence 310 at different moments is encoded using a self-attention model 320 to obtain weights of the route preference features 31-1, 31-2, 31-3…31-n at different moments. The route preference features 31-1, 31-2, 31-3…31-n at different moments are weighted based on the obtained weights and the weighted results are aggregated to obtain the route preference vector of the navigation object representing the route preference of the navigation object. For example, the route preference features 31-1, 31-2, 31-3…31-n are all 300-dimensional vectors, and the user route preference vector can be a 64-dimensional vector, which represents the user's route preference learned from the route preference features 31-1, 31-2, 31-3…31-n. In one embodiment of the present disclosure, the method for obtaining the route preference features 31-1, 31-2, 31-3…31-n can refer to Figure 2 The route preference features of the users shown at different times. In one embodiment of the present disclosure, by aggregating the weighted route preference features, the aggregated results can be added to the target prediction task for the most recent navigation (this task can be selected arbitrarily). By training in the target prediction task, a fully learned self-attention model can be obtained. The learned self-attention model can encode the historical navigation sequence of any user as a navigation object into a single feature vector (user route preference vector). This feature vector represents the user's personalized preference for the route. Although the recommendation method of the prior art can also use the self-attention model for preference learning, the representation method of a single route preference feature is completely different from the embodiment of the present disclosure. Unlike the existing recommendation method that uses pre-training and Embedding technology for representation, the embodiment of the present disclosure directly uses the basic features of the maximum coverage route for representation, which is suitable for obtaining user route preferences through a machine learning model.

[0086] In one embodiment of the present disclosure, the weights of the route preference features at different moments represent the similarities between the route preference features at different moments.

[0087] exist Figure 3 In the route preference feature sequence 310 shown, the weight of each route preference feature 31-1, 31-2, 31-3...31-n at different times represents the similarity between them. In one embodiment of the present disclosure, the higher the similarity between route preference features, the higher the weight of the corresponding route preference feature.

[0088] According to the technical solution provided by the embodiments of the present disclosure, the similarity between the route preference features at different times is represented by the weights of the route preference features at different times. The self-attention mechanism can be used to reduce noise in the historical navigation sequence of the user as the navigation target. This also solves the problem of how to describe the user's personal preferences for routes in the personalized route recommendation scenario.

[0089] In one embodiment of the present disclosure, the navigation route preference acquisition method further includes: acquiring the navigation route with the largest coverage rate among the navigation routes of the same navigation object at the target moment as the target navigation route; acquiring the route characteristics of the target navigation route and the statistical environmental characteristics associated with the route characteristics as the route preference characteristics at the target moment; predicting the route preference of the navigation object based on the route preference vector of the navigation object output by the machine learning model and the route preference characteristics at the target moment to obtain the route preference prediction result of the navigation object; and adjusting the parameters of the machine learning model based on the route preference prediction result of the navigation object.

[0090] like Figure 3 As shown, in the process of training the self-attention model, when training is performed based on the current route preference feature sequence at different moments and the route preference vector of the current navigation object is obtained, the route feature of the navigation route with the greatest coverage of the user's actual route in the user navigation information at the new target moment and the statistical environmental features associated with the route feature are obtained as the route preference feature at the target moment (module 330), and the route preference feature at the target moment is combined with the current route preference vector of the navigation object in the prediction module 340 to perform the prediction task, and the self-attention model 320 is trained using the prediction task in the training module 350. The parameters of the self-attention model 320 can be adjusted according to the training results, and a fully learned self-attention model 320 can be obtained through repeated training.

[0091] According to the technical solution provided by the embodiment of the present disclosure, the navigation route with the largest coverage rate among the navigation routes of the same navigation object at the target moment is obtained as the target navigation route; the route characteristics of the target navigation route and the statistical environmental characteristics associated with the route characteristics are obtained as the route preference characteristics at the target moment; the route preference vector of the navigation object output by the machine learning model and the route preference characteristics at the target moment are used to predict the route preference of the navigation object to obtain the route preference prediction result of the navigation object; the parameters of the machine learning model are adjusted based on the route preference prediction result of the navigation object, and a new route feature representation method can be used to complete route preference learning, thereby truly and comprehensively reflecting the personalized route preference of the navigation object. Moreover, it is helpful to make route recommendations based on the real personalized needs of the user as the navigation object, or to help determine whether the route planned by the navigation meets the real personalized needs of the user as the navigation object.

[0092] The following reference Figure 4 A navigation route preference acquisition device according to an embodiment of the present disclosure is described. Figure 4 FIG. 4 shows a structural block diagram of a navigation route preference acquisition device 400 according to an embodiment of the present disclosure.

[0093] like Figure 4 As shown, the navigation route preference acquisition device 400 includes: a target navigation route acquisition module 401, a route preference feature acquisition module 402, a route preference feature sequence acquisition module 403, and a route preference vector acquisition module 404. The target navigation route acquisition module 401 is configured to acquire the navigation route with the highest coverage rate among the navigation routes of the same navigation object at different times as the target navigation route; the route preference feature acquisition module 402 is configured to acquire the route features of the target navigation route and the statistical environmental features associated with the route features as the route preference features of the target navigation route at the corresponding time; the route preference feature sequence acquisition module 403 is configured to acquire a route preference feature sequence based on the route preference features corresponding to different times; and the route preference vector acquisition module 404 is configured to use the route preference feature sequence as the input of a machine learning model and obtain the route preference vector of the navigation object through the machine learning model.

[0094] According to the technical solution provided by the embodiment of the present disclosure, a target navigation route acquisition module is configured to acquire the navigation route with the largest coverage rate among the navigation routes of the same navigation object at different times as the target navigation route; a route preference feature acquisition module is configured to acquire the route features of the target navigation route and the statistical environmental features associated with the route features as the route preference features of the target navigation route at the corresponding moment; a route preference feature sequence acquisition module is configured to acquire a route preference feature sequence based on the route preference features corresponding to different moments; a route preference vector acquisition module is configured to use the route preference feature sequence as the input of a machine learning model, and obtain the route preference vector of the navigation object through the machine learning model. A new route feature representation method can be used to complete route preference learning, thereby truly and comprehensively reflecting the personalized route preferences of the navigation object. Moreover, it is helpful to make route recommendations based on the real personalized needs of the user as the navigation object, or to help determine whether the route planned by the navigation meets the real personalized needs of the user as the navigation object.

[0095] In one embodiment of the present disclosure, the route preference vector acquisition module 404 is further configured to: encode the route preference feature sequences at different moments using a self-attention model to obtain weights of the route preference features at different moments; weight the route preference features at different moments based on the weights, and aggregate the weighted results to obtain the route preference vector of the navigation object representing the route preference of the navigation object.

[0096] According to the technical solution provided by the embodiment of the present disclosure, the route preference vector acquisition module is further configured to: encode the route preference feature sequences at different times using a self-attention model to obtain weights of the route preference features at different times; weight the route preference features at different times based on the weights, and aggregate the weighted results to obtain the route preference vector of the navigation object representing the route preference of the navigation object. The self-attention mechanism can be used to reduce noise on the historical navigation sequence of the user who is the navigation object, and the problem of how to describe the user's personal preference for routes in the personalized route recommendation scenario is solved.

[0097] In one embodiment of the present disclosure, the weights of the route preference features at different moments represent the similarities between the route preference features at different moments.

[0098] According to the technical solution provided by the embodiments of the present disclosure, the similarity between the route preference features at different times is represented by the weights of the route preference features at different times. The self-attention mechanism can be used to reduce noise in the historical navigation sequence of the user as the navigation target. This also solves the problem of how to describe the user's personal preferences for routes in the personalized route recommendation scenario.

[0099] In one embodiment of the present disclosure, the navigation route preference acquisition device further includes: a target moment route preference feature acquisition module, configured to acquire the navigation route with the largest coverage rate among the navigation routes of the same navigation object at the target moment as the target navigation route; a target moment route preference feature acquisition module, configured to acquire the route features of the target navigation route and the statistical environmental features associated with the route features as the route preference features at the target moment; a route preference prediction result acquisition module, configured to perform route preference prediction of the navigation object based on the route preference vector of the navigation object output by the machine learning model and the route preference features at the target moment, so as to acquire the route preference prediction result of the navigation object; a parameter adjustment module, configured to adjust the parameters of the machine learning model based on the route preference prediction result of the navigation object, so as to acquire the route preference prediction result of the navigation object; and a parameter adjustment module, configured to adjust the parameters of the machine learning model based on the route preference prediction result of the navigation object.

[0100] According to the technical solution provided by the embodiment of the present disclosure, a target moment route preference feature acquisition module is configured to acquire the navigation route with the highest coverage rate among the navigation routes of the same navigation object at the target moment as the target navigation route; a target moment route preference feature acquisition module is configured to acquire the route features of the target navigation route and the statistical environmental features associated with the route features as the route preference features of the target moment; a route preference prediction result acquisition module is configured to predict the route preference of the navigation object based on the route preference vector of the navigation object output by the machine learning model and the route preference features at the target moment, so as to obtain the route preference prediction result of the navigation object; a parameter adjustment module is configured to adjust the parameters of the machine learning model based on the route preference prediction result of the navigation object to obtain the route preference prediction result of the navigation object; and a parameter adjustment module is configured to adjust the parameters of the machine learning model based on the route preference prediction result of the navigation object. A new route feature representation method can be used to complete route preference learning, thereby truly and comprehensively reflecting the personalized route preferences of the navigation object. Moreover, it is helpful to make route recommendations based on the real personalized needs of the user as the navigation object, or to help determine whether the route planned by the navigation object meets the real personalized needs of the user as the navigation object.

[0101] In one embodiment of the present disclosure, the route feature is a route feature vector including multidimensional route features, the statistical environment feature is a statistical environment feature vector including multidimensional statistical environment features, and the route preference feature includes the route feature vector and the statistical environment feature vector, wherein the statistical environment feature vector includes the maximum eigenvalue and the minimum eigenvalue of each dimensional feature in the route feature vector among the route features of the corresponding dimensions of all navigation routes.

[0102] According to the technical solution provided by the embodiment of the present disclosure, the route feature is a route feature vector including multi-dimensional route features, the statistical environment feature is a statistical environment feature vector including multi-dimensional statistical environment features, and the route preference feature includes the route feature vector and the statistical environment feature vector, wherein the statistical environment feature vector includes the maximum eigenvalue and the minimum eigenvalue of the route features of the corresponding dimensions of all navigation routes in the route feature vector for each dimensional feature, a new route feature representation method can be used to complete route preference learning, thereby truly and comprehensively reflecting the personalized route preference of the navigation object.

[0103] Those skilled in the art will understand that referring to Figure 4 The technical solutions described can be compared with Figures 1 to 3 The embodiments described are combined to provide a reference Figures 1 to 3 The technical effects achieved by the described embodiments can be referred to above for details. Figures 1 to 3 The specific content will not be repeated here.

[0104] In one embodiment of the present disclosure, a navigation route recommendation method is provided, the method comprising:

[0105] Get the navigation route to be recommended to the navigation object;

[0106] determining a route preference vector of the navigation object;

[0107] Based on the route preference vector of the navigation object, a recommended order of the navigation routes is determined, wherein the route preference vector of the navigation object is determined by the aforementioned navigation route preference acquisition method.

[0108] For example, the route preference vector of the navigation object is determined by the following steps:

[0109] Obtain the navigation route with the largest coverage rate among the navigation routes of the same navigation object at different times as the target navigation route;

[0110] Acquiring route characteristics of the target navigation route and statistical environmental characteristics associated with the route characteristics as route preference characteristics at a corresponding moment of the target navigation route;

[0111] obtaining a route preference feature sequence based on the route preference features corresponding to different moments;

[0112] The route preference feature sequence is used as an input of a machine learning model, and a route preference vector of the navigation object is obtained through the machine learning model.

[0113] In one embodiment of the present disclosure, a fully learned machine learning model, such as a self-attention model, can be used to obtain a user route preference vector. Based on the user route preference vector, the navigation route recommended to the user will be consistent with the user's actual preferences. Those skilled in the art will appreciate that this recommendation method can be used independently or in combination with other recommendation methods.

[0114] In addition, the user route preference vector obtained using the fully learned machine learning model can be used to evaluate the degree to which the navigation route recommended to the user complies with the user's preferences.

[0115] According to the technical solution provided by the embodiment of the present disclosure, a navigation route to be recommended to a navigation object is obtained; a route preference vector of the navigation object is determined; and a recommendation order of the navigation route is determined based on the route preference vector of the navigation object. The route preference vector of the navigation object is determined by the aforementioned navigation route preference acquisition method. A new route feature representation method can be used to complete route preference learning, thereby truly and comprehensively reflecting the personalized route preferences of the navigation object. Furthermore, route recommendations can be made based on the actual personalized needs of the user who is the navigation object, or it can be determined whether the route planned by the navigation object meets the actual personalized needs of the user who is the navigation object. A new route feature representation method can be used to complete route preference learning, thereby truly and comprehensively reflecting the personalized route preferences of the navigation object. Furthermore, route recommendations can be made based on the actual personalized needs of the user who is the navigation object.

[0116] In one embodiment of the present disclosure, a navigation route recommendation device is provided, the device comprising:

[0117] A navigation route acquisition module is configured to acquire a navigation route to be recommended to a navigation object;

[0118] The navigation route preference acquisition device in the aforementioned embodiment is configured to determine the route preference vector of the navigation object;

[0119] The recommendation order determination module is configured to determine the recommendation order of the navigation routes based on the route preference vector of the navigation object, wherein the route preference vector of the navigation object is determined by the navigation route preference acquisition device in the aforementioned embodiment.

[0120] For example, the navigation route preference acquisition device includes:

[0121] A target navigation route acquisition module is configured to acquire the navigation route with the largest coverage rate among the navigation routes of the same navigation object at different times as the target navigation route;

[0122] a route preference feature acquisition module configured to acquire the route features of the target navigation route and the statistical environment features associated with the route features as the route preference features of the target navigation route at a corresponding moment;

[0123] a route preference feature sequence acquisition module, configured to acquire a route preference feature sequence based on the route preference features corresponding to different moments;

[0124] The route preference vector acquisition module is configured to use the route preference feature sequence as an input of a machine learning model, and obtain the route preference vector of the navigation object through the machine learning model.

[0125] According to the technical solution provided by the embodiments of the present disclosure, a navigation route acquisition module is configured to acquire a navigation route to be recommended to a navigation subject; the navigation route preference acquisition device in the aforementioned embodiments is configured to determine the route preference vector of the navigation subject; and the recommendation order determination module is configured to determine the recommendation order of the navigation routes based on the route preference vector of the navigation subject. Determining the route preference vector of the navigation subject by the navigation route preference acquisition device in the aforementioned embodiments allows for a new route feature representation method to be used to complete route preference learning, thereby truly and comprehensively reflecting the personalized route preferences of the navigation subject. Furthermore, route recommendations can be made based on the actual personalized needs of the user serving as the navigation subject.

[0126] In particular, according to an embodiment of the present disclosure, the method described above with reference to the accompanying drawings can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product comprising computer instructions, which implement the method as shown in the accompanying drawings when the computer instructions are executed by a processor. The computer program product may include a computer program tangibly contained on a readable medium thereof, the computer program comprising program code for executing the method in the accompanying drawings. In such an embodiment, the computer program can be downloaded and installed from a network via a communication part, and / or installed from a removable medium. For example, an embodiment of the present disclosure includes a readable storage medium having computer instructions stored thereon, which implement the program code for executing the method in the accompanying drawings when the computer instructions are executed by a processor.

[0127] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each box in the diagram or block diagram can represent a module, program segment or part of the code, and the module, program segment or part of the code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flow chart, as well as the combination of boxes in the block diagram and / or flow chart, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or can be implemented using a combination of dedicated hardware and computer instructions.

[0128] The units or modules described in the embodiments of the present disclosure may be implemented in software or hardware. The units or modules described may also be provided in a processor, and the names of these units or modules do not, in certain circumstances, limit the units or modules themselves.

[0129] As another aspect, the present disclosure further provides a computer-readable storage medium. This computer-readable storage medium may be included in the node described in the above embodiments, or may be a standalone computer-readable storage medium not incorporated into a device. The computer-readable storage medium stores one or more programs, which are used by one or more processors to execute the methods described in the present disclosure.

[0130] The above description is merely a preferred embodiment of the present disclosure and an illustration of the technical principles employed. Those skilled in the art should understand that the scope of the invention herein is not limited to the technical solutions formed by the specific combination of the above-mentioned technical features, but also encompasses other technical solutions formed by any combination of the above-mentioned technical features or their equivalents without departing from the inventive concept. For example, a technical solution formed by replacing the above-mentioned features with (but not limited to) technical features with similar functions disclosed in this disclosure.

Claims

1. A method for obtaining navigation route preferences, wherein: The method comprises: Obtaining the navigation route with the largest actual coverage rate among the navigation routes of the same navigation object at different times as the target navigation route, wherein the actual coverage rate is calculated by dividing the length of the overlapping portion of the navigation route and the actual route of the same navigation object by the length of the actual route; Acquiring route characteristics of the target navigation route and statistical environmental characteristics associated with the route characteristics as route preference characteristics at a corresponding moment of the target navigation route; obtaining a route preference feature sequence based on the route preference features corresponding to different moments; The route preference feature sequence is used as an input of a machine learning model, and a route preference vector of the navigation object is obtained through the machine learning model.

2. The method according to claim 1, wherein The step of using the route preference feature sequence as an input to a machine learning model and obtaining a route preference vector of the navigation object through the machine learning model includes: Encoding the route preference feature sequences at different moments using a self-attention model to obtain weights of the route preference features at different moments; The route preference features at different moments are weighted based on the weights, and the weighted results are aggregated to obtain a route preference vector of the navigation object representing the route preference of the navigation object.

3. The method according to claim 2, wherein: The weights of the route preference features at different times represent the similarities between the route preference features at different times.

4. The method according to any one of claims 1 to 3, wherein: The method further comprises: Obtaining the navigation route with the largest coverage rate among the navigation routes of the same navigation object at the target time as the target navigation route; Acquiring route characteristics of the target navigation route and statistical environmental characteristics associated with the route characteristics as route preference characteristics at the target moment; performing a route preference prediction of the navigation object based on the route preference vector of the navigation object output by the machine learning model and the route preference feature at the target time, to obtain a route preference prediction result of the navigation object; Adjust the parameters of the machine learning model based on the route preference prediction result of the navigation object.

5. The method according to any one of claims 1 to 3, wherein: The route feature is a route feature vector including multi-dimensional route features, the statistical environment feature is a statistical environment feature vector including multi-dimensional statistical environment features, and the route preference feature includes the route feature vector and the statistical environment feature vector. The statistical environment feature vector includes the maximum feature value and the minimum feature value of each dimensional feature in the route feature vector among the route features of the corresponding dimensions of all planned routes.

6. A navigation route preference acquisition device, wherein: The device comprises: a target navigation route acquisition module configured to acquire, as the target navigation route, a navigation route with the largest actual walking coverage rate among navigation routes of the same navigation object at different times, wherein the actual walking coverage rate is calculated as the length of the overlapping portion between the navigation route and the actual walking route of the same navigation object divided by the length of the actual walking route; a route preference feature acquisition module configured to acquire the route features of the target navigation route and the statistical environment features associated with the route features as the route preference features of the target navigation route at a corresponding moment; a route preference feature sequence acquisition module, configured to acquire a route preference feature sequence based on the route preference features corresponding to different moments; The route preference vector acquisition module is configured to use the route preference feature sequence as an input of a machine learning model, and obtain the route preference vector of the navigation object through the machine learning model.

7. The device according to claim 6, wherein The route preference vector acquisition module is further configured to: Encoding the route preference feature sequences at different moments using a self-attention model to obtain weights of the route preference features at different moments; The route preference features at different moments are weighted based on the weights, and the weighted results are aggregated to obtain a route preference vector of the navigation object representing the route preference of the navigation object.

8. A navigation route recommendation method, wherein: The method comprises: Get the navigation route to be recommended to the navigation object; determining a route preference vector of the navigation object; Based on the route preference vector of the navigation object, a recommended order of the navigation routes is determined, wherein the route preference vector of the navigation object is determined by the method of any one of claims 1-5.

9. A navigation route recommendation device, wherein: The device comprises: A navigation route acquisition module is configured to acquire a navigation route to be recommended to a navigation object; The navigation route preference acquisition device of claim 6 or 7, configured to determine a route preference vector of the navigation object; The recommended order determination module is configured to determine the recommended order of the navigation routes based on the route preference vector of the navigation object, wherein the route preference vector of the navigation object is determined by the navigation route preference acquisition device of claim 6 or 7.

10. A computer program product comprising computer instructions, wherein when the computer instructions are executed by a processor, the method according to any one of claims 1 to 5 and 8 is implemented.

Images