Route recommendation method and device, electronic equipment and storage medium

Abstract

The application discloses a route recommendation method and device, electronic equipment and a storage medium. The route recommendation method comprises the following steps: determining at least one candidate route from a target starting point to a target ending point; determining candidate route features of the candidate route according to candidate route data of the at least one candidate route; determining a target route from the at least one candidate route according to the candidate route features based on a route recommendation model, and recommending the target route to a demand party. The above technical solution solves the problem of unsatisfactory route recommendation effect in the prior art, greatly improves the accuracy of route recommendation, and is more in line with the travel preferences of users.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of sorting algorithm and machine learning algorithm, and particularly relates to a route recommendation method and device, equipment and a storage medium. BACKGROUND

[0002] When calculating a route, navigation will calculate multiple routes, and consider actively recommending an optimal route to the user, and other routes as backup routes for the user to choose. The existing method is more of a sorting design based on a certain user preference assumption, which does not necessarily truly meet the user's preference needs.

[0003] The existing algorithm mainly sorts or calculates a comprehensive weight by assigning a weight coefficient to multiple dimensions from the route travel time (ETA time), route distance, number of traffic lights on the route, and route toll amount. Among them, single-dimensional sorting cannot achieve the optimal sorting result because the user's selection of the route is a comprehensive selection of multiple factors; and multi-dimensional assignment of a weight coefficient to calculate a comprehensive weight to achieve route sorting is difficult to define the most reasonable weight coefficient of each dimension to make the final sorting result truly meet the user's preference. It has been verified that there is not a simple linear relationship between the dimensions. Therefore, there is an urgent need for a route recommendation method that meets the user's preference. SUMMARY

[0004] The present application provides a route recommendation method, device, equipment and storage medium to achieve accurate recommendation of user travel routes.

[0005] In a first aspect, a route recommendation method is provided, which comprises:

[0006] According to the candidate route data of at least one candidate route, the candidate route characteristics of the candidate route are determined; wherein the candidate route characteristics include candidate navigation characteristics and / or candidate road characteristics;

[0007] According to the candidate route characteristics, a target route is determined from at least one candidate route based on a route recommendation model, and the target route is recommended to the demand side; wherein the route recommendation model is a model obtained by pre-training an integrated classifier.

[0008] In a second aspect, a training method of a route recommendation model is provided, which comprises:

[0009] At least one candidate route from a target starting point to a target ending point is determined;

[0010] Based on candidate route data from at least one candidate route, candidate route features are determined; wherein, the candidate route features include candidate navigation features and / or candidate road features;

[0011] The route recommendation model is trained based on the candidate route features and the ground truth penalty values ​​of the candidate routes; wherein, the route recommendation model is a model obtained by pre-training an ensemble classifier.

[0012] Thirdly, a route recommendation device is provided, the device comprising:

[0013] A candidate route feature determination module is used to determine candidate route features of the candidate route based on candidate route data of at least one candidate route; wherein, the candidate route features include candidate navigation features and / or candidate road features;

[0014] The target route recommendation module is used to determine a target route from at least one candidate route based on the characteristics of the candidate routes using a route recommendation model, and recommend the target route to the demander; wherein, the route recommendation model is a model obtained by training an ensemble classifier.

[0015] Fourthly, a training device for a route recommendation model is provided, the device comprising:

[0016] The candidate route determination module is used to determine at least one candidate route from the target starting point to the target ending point;

[0017] A candidate route feature determination module is used to determine candidate route features of the candidate route based on candidate route data of at least one candidate route; wherein, the candidate route features include candidate navigation features and / or candidate road features;

[0018] The route recommendation model training module trains the route recommendation model based on the candidate route features and the ground truth penalty values ​​of the candidate routes; wherein the route recommendation model is a model obtained by pre-training an ensemble classifier.

[0019] Fifthly, embodiments of the present invention also provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the route recommendation method or the route recommendation model training method as described in any of the embodiments of the present invention.

[0020] Sixthly, embodiments of the present invention also provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the route recommendation method or route recommendation model training method as described in any of the embodiments of the present invention.

[0021] The technical solution of this invention, by determining the candidate route features of at least one candidate route and inputting the candidate route features of at least one candidate route into a route recommendation model, determines a target route from at least one candidate route and recommends it to the user. This solves the problem of unsatisfactory route recommendation results in the prior art, greatly improves the accuracy of route recommendation, and is more in line with the user's travel preferences.

[0022] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1A This is a flowchart of a route recommendation method provided in Embodiment 1 of the present invention;

[0025] Figure 1B This is a schematic diagram of the XGBoosting algorithm parameter settings provided in Embodiment 1 of the present invention;

[0026] Figure 2 This is a flowchart of a route recommendation method provided in Embodiment 2 of the present invention;

[0027] Figure 3 This is a flowchart of a training method for a route recommendation model according to Embodiment 3 of the present invention;

[0028] Figure 4 This is a schematic diagram of a route recommendation device according to Embodiment 4 of the present invention;

[0029] Figure 5 This is a schematic diagram of a training device for a route recommendation model according to Embodiment 5 of the present invention;

[0030] Figure 6 This is a schematic diagram of the structure of an electronic device that implements the route recommendation method or the training method of the route recommendation model in the embodiments of the present invention. Detailed Implementation

[0031] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0032] It should be noted that the terms "target," "candidate," etc., used in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0033] Furthermore, it should be noted that the collection, storage, use, processing, transmission, provision, and disclosure of candidate route data and other related data in the technical solution of this invention all comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0034] Example 1

[0035] Figure 1A The flowchart illustrates a route recommendation method provided in Embodiment 1 of the present invention. This embodiment is applicable to situations requiring accurate route recommendations to users. The method can be executed by a route recommendation device, which can be implemented in hardware and / or software and integrated into an electronic device carrying route recommendation functionality, such as a server. Figure 1A As shown, the method includes:

[0036] S110. Determine at least one candidate route from the target starting point to the target ending point.

[0037] The starting point can refer to the starting address entered by the user in real time on the navigation device. The destination can refer to the desired destination address entered by the user in real time on the navigation device. Optionally, the navigation device can refer to a mobile device with positioning capabilities, such as a mobile phone or in-vehicle terminal.

[0038] Among them, the candidate route can refer to the feasible route from the target starting point to the target destination provided by the navigation device after the user enters the target starting point and target destination.

[0039] Specifically, after a user with navigation needs inputs their target origin and destination into the navigation device, the device will provide at least one feasible route from the origin to the destination as candidate routes. For example, if a user wants to travel from origin A to destination B, after inputting origin A and destination B into the navigation device, the device will provide three optional routes from origin A to destination B as candidate routes.

[0040] S120. Based on the candidate route data of at least one candidate route, determine the candidate route characteristics of the candidate route.

[0041] Candidate route data can refer to route information within a candidate route. Optionally, candidate route data may include candidate navigation data and / or candidate road data. For example, candidate navigation data may refer to navigation information for a candidate route provided by a navigation device; optionally, candidate navigation data may include at least one of the following: time data, distance data, cost data, traffic light data, and turning data for the candidate route. For example, candidate road data includes first road data and the total length of the candidate route; the first road data includes the segment length of the candidate road level (expressway, national highway, provincial highway, etc.) and / or candidate road condition data; wherein, candidate road condition data refers to the traffic flow status of the road, which may include congestion length or smooth flow length, etc.

[0042] For example, as shown in Table 1, Table 1 records the candidate route data for three candidate routes from a target starting point to a target ending point. Optionally, time ( / minute) can refer to the time required for the candidate route, distance ( / m) can refer to the distance of the candidate route, amount ( / yuan) can refer to the amount spent on the candidate route, traffic lights ( / number) can refer to the traffic light data in the candidate route, highway ( / m) can refer to the length of the highways passed by in the candidate route, national highway ( / m) can refer to the length of the national highways passed by in the candidate route, congestion ( / m) can refer to the road condition data in the candidate route, and left turns ( / number) can refer to the turning data in the candidate route.

[0043] Table 1

[0044]

[0045] The so-called candidate route features (referring to the features obtained after processing candidate route data) are the characteristics of candidate routes, which can be represented in the form of vectors or matrices; optionally, candidate route features include candidate navigation features and / or candidate road features. Specifically, candidate navigation features refer to the features obtained after processing candidate navigation data, which can be represented in matrix or vector form; candidate road features refer to the features obtained after processing candidate road data, which can also be represented in matrix or vector form.

[0046] Specifically, based on a feature extraction model, candidate route features can be determined from candidate route data. This feature extraction model can be composed of a convolutional neural network and pre-trained.

[0047] S130. Based on the route recommendation model, determine the target route from at least one candidate route according to the characteristics of the candidate routes, and recommend the target route to the demander.

[0048] The route recommendation model can be a model obtained by pre-training an ensemble classifier.

[0049] Here, the ensemble classifier can refer to the XGBoosting algorithm. Furthermore, the parameters of the XGBoosting algorithm need to be reset. For example, the parameter settings for the XGBoosting algorithm can be as follows: Figure 1B As shown. The `booster` parameter refers to the model setting of the XGBoosting algorithm. In this embodiment, the XGBoosting model is preferably set to `gbtree`: a tree-based model. The `objective` parameter defines the loss function to be minimized. In this embodiment, at least one candidate path needs to be classified and predicted, so the `objective` parameter is preferably set to `multi:softmax`. The `num_class` parameter represents the number of class parameters. Preferably, in this embodiment, three candidate paths can be predicted, and `num_class` can be set to 3. The `max_depth` parameter refers to the maximum depth of the tree. Preferably, to prevent overfitting, `max_depth` can be set to 6 in this embodiment. The `eta` parameter is the learning rate of the value model. Preferably, to improve the stability of the trained model, `eta` can be set to 0.01 in this embodiment. The `gamma` parameter is a parameter used to optimize the model and reduce model redundancy. Optimized, `gamma` can be set to 0.1 in this embodiment. The `lambda` parameter is a regularization term parameter that controls the model complexity of the weight values. Preferably, `lambda` can be set to 2 in this embodiment.

[0050] The target route can refer to the route ultimately recommended to the user. The demand side can refer to the party requesting specific route recommendations, such as the party operating the navigation device.

[0051] Specifically, candidate route features can be input into the route recommendation model. The route recommendation model will process the candidate route features and output the route that best matches the user's actual choice from at least one candidate route as the target route to recommend to the user.

[0052] Optionally, based on the route recommendation model, the target route is determined from at least one candidate route according to the characteristics of the candidate routes, including: based on the route recommendation model, determining the route penalty score of the candidate routes according to the characteristics of the candidate routes; ranking the at least one candidate route according to the route penalty score; and determining the target route according to the ranking result.

[0053] The route penalty score characterizes the degree of matching between the candidate route and the user's actual walking route; that is, the degree of matching between the candidate route predicted by the route recommendation model and the user's actual walking route. For example, if there are three candidate routes, the candidate route with the highest matching degree has a penalty of 0, the candidate route with the second highest matching degree has a penalty of 1, and the candidate route with the lowest matching degree has a penalty of 2.

[0054] Specifically, based on the route recommendation model, candidate route features can be input into the route recommendation model, which will output the route penalty score of the candidate route corresponding to the candidate route feature. According to the obtained route penalty scores of the candidate routes, at least one candidate route is sorted in ascending (or descending) order. Based on the sorting results, the candidate route with the lowest route penalty score is selected as the target route and recommended to the user, while the remaining candidate routes can be used as alternative routes.

[0055] By sorting the route penalty scores of candidate routes output by the route recommendation model, the target route can be quickly determined from the candidate routes and recommended to the user, thus improving the efficiency of target route determination.

[0056] The technical solution of this invention, by determining the candidate route features of at least one candidate route and inputting the candidate route features of at least one candidate route into a route recommendation model, determines a target route from at least one candidate route and recommends it to the user. This solves the problem of unsatisfactory route recommendation results in the prior art, greatly improves the accuracy of route recommendation, and is more in line with the user's travel preferences.

[0057] Example 2

[0058] Figure 2This is a flowchart of a route recommendation method provided in Embodiment 2 of the present invention. Based on the above embodiments, this embodiment further optimizes the determination of candidate route features based on candidate route data from at least one candidate path. For example... Figure 2 As shown, the method includes:

[0059] S210. Determine at least one candidate route from the target starting point to the target ending point.

[0060] S220. Based on the candidate navigation data in the candidate route data of at least one candidate route, determine the mean value of the candidate navigation data of at least one candidate route.

[0061] The mean of candidate navigation data can refer to the average value of candidate navigation data. Optionally, the mean of candidate navigation data may include at least one of the following: the mean of time data for candidate routes, the mean of distance data for candidate routes, the mean of monetary data for candidate routes, the mean of traffic light data for candidate routes, and the mean of turning data for candidate routes.

[0062] Specifically, based on at least one candidate navigation data included in the candidate route data of at least one candidate route, the same type of candidate navigation data in the at least one candidate route is summed separately, and then the average value of the candidate navigation data is obtained based on the number of candidate routes.

[0063] S230. Based on the mean of the candidate navigation data and the candidate navigation data of the candidate routes, determine the candidate navigation features of the candidate routes.

[0064] The candidate navigation features may include at least one of the following: candidate time features, candidate route features, candidate amount features, candidate traffic light features, and candidate turning features.

[0065] Specifically, based on the mean of candidate navigation data and the candidate navigation data of candidate routes, the difference between the navigation data of candidate routes and the mean of candidate navigation data is determined as the candidate navigation feature of the candidate routes.

[0066] For example, as shown in Table 2, Table 2 records the candidate time characteristics, candidate distance characteristics, candidate amount characteristics, candidate traffic light characteristics, and candidate turning characteristics of three candidate routes from a target starting point to a target ending point.

[0067] For example, combining the candidate route data recorded in Table 1, the candidate time characteristic of candidate route 1 is 31.25 - (31.25 + 35.45 + 30.7) ÷ 3, the candidate time characteristic of candidate route 2 is 35.45 - (31.25 + 35.45 + 30.7) ÷ 3, and the candidate time characteristic of candidate route 3 is 30.7 - (31.25 + 35.45 + 30.7) ÷ 3.

[0068] Candidate route characteristics of candidate route 1 = 21718 - (21718 + 21313 + 26496) ÷ 3; Candidate route data of candidate route 2 = 21313 - (21718 + 21313 + 26496) ÷ 3; Candidate route data of candidate route 3 = 26496 - (21718 + 21313 + 26496) ÷ 3.

[0069] Candidate amount characteristics of candidate route 1 = 0 - (0 + 0 + 10) ÷ 3, candidate amount characteristics of candidate route 2 = 0 - (0 + 0 + 10) ÷ 3, candidate amount characteristics of candidate route 3 = 10 - (0 + 0 + 10) ÷ 3.

[0070] The candidate traffic light characteristics for candidate route 1 are 18 - (18 + 20 + 7) ÷ 3, the candidate traffic light characteristics for candidate route 2 are 20 - (18 + 20 + 7) ÷ 3, and the candidate traffic light characteristics for candidate route 3 are 7 - (18 + 20 + 7) ÷ 3.

[0071] Candidate turning characteristics of candidate route 1 = 3 - (3 + 3 + 2) ÷ 3, candidate turning characteristics of candidate route 2 = 3 - (3 + 3 + 2) ÷ 3, candidate turning characteristics of candidate route 3 = 2 - (3 + 3 + 2) ÷ 3.

[0072] Optionally, the candidate navigation features of the three candidate routes recorded in Table 2 can better express the relative features of one candidate route and the remaining routes in the same candidate navigation data, relative to the candidate route data of the candidate routes.

[0073] Table 2

[0074]

[0075] S240. Determine candidate road data from the candidate route data of at least one candidate route.

[0076] The candidate road data may include first road data and the total length of the candidate routes. The first road data may include the segment length of the candidate road level and / or candidate road condition data. The total length of the candidate routes is the distance data of the candidate routes.

[0077] Specifically, candidate road data included in the candidate route data is determined from the candidate route data of the candidate routes.

[0078] S250. Based on the first road data and the total length of the candidate routes, determine the first proportion of the candidate routes.

[0079] The first percentage can refer to the ratio of the first road data to the total length of the candidate routes.

[0080] Specifically, based on the first road data and the total length of the candidate routes, the ratio of the first road data to the total length of the candidate routes is determined as the first proportion of the first road data in the candidate road data of the candidate routes.

[0081] S260. Based on the first percentage of candidate routes, determine the average percentage of at least one candidate route.

[0082] Here, the average percentage can refer to the average of the first percentage.

[0083] Specifically, based on the first proportion of at least one type of first road data included in the candidate road data of at least one candidate route, the first proportions of the same type of first road data in at least one candidate route are summed, and then the average value of the first proportions is obtained based on the number of candidate routes.

[0084] S270. Based on the average percentage and the first percentage, determine the candidate road characteristics of the candidate routes.

[0085] Candidate road features may include candidate road grade features and / or candidate road condition features.

[0086] Specifically, based on the average proportion of candidate routes and the first proportion, the difference between the first proportion of candidate routes and the average proportion is determined as the candidate road feature of the candidate route.

[0087] In another optional embodiment of the present invention, the first proportion of the candidate routes can be directly determined as the candidate road characteristics of the candidate routes. For example, as shown in Table 3, Table 3 records the candidate road level characteristics and candidate road condition characteristics of three candidate routes from a target starting point to a target ending point.

[0088] For example, taking the candidate route data recorded in Table 1 as an example, the candidate road characteristics of candidate route 1 can include candidate expressway characteristics and candidate national highway characteristics. The candidate expressway characteristics of candidate route 1 = 1252 ÷ 21718, the candidate expressway characteristics of candidate route 2 = 1058 ÷ 21313, and the candidate expressway characteristics of candidate route 3 = 1389 ÷ 26496.

[0089] Candidate national highway grade characteristics of candidate route 1 = 474 ÷ 21718, candidate passage grade characteristics of candidate route 2 = 546 ÷ 21313, candidate national highway grade of candidate route 3 = 758 ÷ 26496.

[0090] For example, combining the candidate route data recorded in Table 1, and determining the candidate road characteristics of the candidate routes based on the average proportion and the first proportion, the candidate road condition characteristics of candidate route 1 are as follows: 2011÷21718-[(2011÷21718)+(1856÷21313)+(1674÷26496)]÷3; the candidate road condition data of candidate route 2 are as follows: 1856÷21313-[(2011÷21718)+(1856÷21313)+(1674÷26496)]÷3; the candidate road condition characteristics of candidate route 3 are as follows: 1674÷26496-[(2011÷21718)+(1856÷21313)+(1674÷26496)]÷3.

[0091] Optionally, the candidate road features of the three candidate routes recorded in Table 3 can better express the relative features of one candidate route and the same candidate road data of the remaining routes compared to the candidate route data of the candidate routes.

[0092] Table 3

[0093] ID candidate highway grade features candidate national road grade features candidate road condition features 1 0.058 0.021 0.012 2 0.050 0.026 0.006 3 0.052 0.029 -0.018

[0094] In this invention, by extracting candidate navigation features and / or candidate route features from the candidate path information of at least one candidate route and using them as input to the route recommendation model, the route recommendation model can identify the relative features of route data between different candidate routes, better distinguish the differences in candidate route information between different candidate routes, improve the matching degree between the target route output by the route recommendation model and the user, and improve the working efficiency of the route recommendation model.

[0095] Example 3

[0096] Figure 3 This is a flowchart illustrating a training method for a route recommendation model provided in Embodiment 3 of the present invention. This embodiment is applicable to situations where accurate route recommendations are provided to users. This method can be executed by a route recommendation model training device, which can be implemented in hardware and / or software and integrated into an electronic device that carries the training function of the route recommendation model, such as a server. Figure 3 As shown, the method includes:

[0097] S310. Determine at least one candidate route from the target starting point to the target ending point.

[0098] The starting point can refer to the starting address historically entered by the user on the navigation device. The destination can refer to the desired destination address historically entered by the user on the navigation device. Optionally, the navigation device can refer to a mobile device with positioning capabilities, such as a mobile phone or in-vehicle terminal.

[0099] Among them, the candidate route can refer to the feasible route from the target starting point to the target destination provided by the navigation device after the user enters the target starting point and target destination.

[0100] Specifically, based on the target origin and destination historically entered by the user on the navigation device, at least one feasible route from the target origin to the target destination is determined as a candidate route. For example, if a user wants to travel from target origin A to target destination B, after entering target origin A and target destination B into the navigation device, the three optional paths from target origin A to target destination B provided by the navigation device are selected as candidate routes.

[0101] S320. Based on the candidate route data of at least one candidate route, determine the candidate route characteristics of the candidate route.

[0102] Candidate route data can refer to route information within a candidate route. Optionally, candidate route data may include candidate navigation data and / or candidate road data. For example, candidate navigation data may refer to navigation information for a candidate route provided by a navigation device; optionally, candidate navigation data may include at least one of the following: time data, distance data, cost data, traffic light data, and turning data for the candidate route. For example, candidate road data includes first road data and the total length of the candidate route; the first road data includes the segment length of the candidate road level (expressway, national highway, provincial highway, etc.) and / or candidate road condition data; wherein, candidate road condition data refers to the traffic flow status of the road, which may include congestion length or smooth flow length, etc.

[0103] For example, as shown in Table 4, Table 4 records the candidate route data for three candidate routes from a target starting point to a target ending point. Optionally, time ( / minute) can refer to the time required for the candidate route, distance ( / m) can refer to the distance of the candidate route, amount ( / yuan) can refer to the amount spent on the candidate route, traffic lights ( / number) can refer to the traffic light data in the candidate route, highway ( / m) can refer to the length of the highways passed by in the candidate route, national highway ( / m) can refer to the length of the national highways passed by in the candidate route, congestion ( / m) can refer to the road condition data in the candidate route, and left turns ( / number) can refer to the turning data in the candidate route.

[0104] Table 4

[0105]

[0106] Candidate route features refer to the features obtained after processing candidate route data, which can be represented in the form of vectors or matrices. Optionally, candidate route features include candidate navigation features and / or candidate road features. Candidate navigation features refer to the features obtained after processing candidate navigation data, which can be represented in the form of matrices or vectors; candidate road features refer to the features obtained after processing candidate road data, which can be represented in the form of matrices or vectors.

[0107] Specifically, based on a feature extraction model, candidate route features can be determined from candidate route data. This feature extraction model can be composed of a convolutional neural network and pre-trained.

[0108] Optionally, based on the candidate route data of at least one candidate route, the candidate route features of the candidate route are determined, including: determining the mean of the candidate navigation data of at least one candidate route based on the candidate navigation data in the candidate route data of at least one candidate route; and determining the candidate navigation features of the candidate route based on the mean of the candidate navigation data and the candidate navigation data of the candidate route.

[0109] The mean of candidate navigation data can refer to the average value of candidate navigation data. Optionally, the mean of candidate navigation data may include at least one of the following: the mean of time data for candidate routes, the mean of distance data for candidate routes, the mean of monetary data for candidate routes, the mean of traffic light data for candidate routes, and the mean of turning data for candidate routes.

[0110] Specifically, based on at least one candidate navigation data included in the candidate route data of at least one candidate route, the same type of candidate navigation data in the at least one candidate route is summed separately, and then the average value of the candidate navigation data is obtained based on the number of candidate routes.

[0111] The candidate navigation features may include at least one of the following: candidate time features, candidate route features, candidate amount features, candidate traffic light features, and candidate turning features.

[0112] Specifically, based on the mean of candidate navigation data and the candidate navigation data of candidate routes, the difference between the navigation data of candidate routes and the mean of candidate navigation data is determined as the candidate navigation feature of the candidate routes.

[0113] For example, as shown in Table 5, Table 5 records the candidate time characteristics, candidate distance characteristics, candidate amount characteristics, candidate traffic light characteristics, and candidate turning characteristics of three candidate routes from a target starting point to a target ending point.

[0114] For example, combining the candidate route data in Table 4, the candidate time characteristic of candidate route 1 is 31.25 - (31.25 + 35.45 + 30.7) ÷ 3. The candidate time characteristic of candidate route 2 is 35.45 - (31.25 + 35.45 + 30.7) ÷ 3. The candidate time characteristic of candidate route 3 is 30.7 - (31.25 + 35.45 + 30.7) ÷ 3.

[0115] Candidate route 1's candidate route characteristics = 21718 - (21718 + 21313 + 26496) ÷ 3. Candidate route 2's candidate route data = 21313 - (21718 + 21313 + 26496) ÷ 3. Candidate route 3's candidate route data = 26496 - (21718 + 21313 + 26496) ÷ 3.

[0116] Candidate amount characteristics of candidate route 1 = 0 - (0 + 0 + 10) ÷ 3, candidate amount characteristics of candidate route 2 = 0 - (0 + 0 + 10) ÷ 3, candidate amount characteristics of candidate route 3 = 10 - (0 + 0 + 10) ÷ 3.

[0117] The candidate traffic light characteristics for candidate route 1 are 18 - (18 + 20 + 7) ÷ 3, the candidate traffic light characteristics for candidate route 2 are 20 - (18 + 20 + 7) ÷ 3, and the candidate traffic light characteristics for candidate route 3 are 7 - (18 + 20 + 7) ÷ 3.

[0118] Candidate turning characteristics of candidate route 1 = 3 - (3 + 3 + 2) ÷ 3, candidate turning characteristics of candidate route 2 = 3 - (3 + 3 + 2) ÷ 3, candidate turning characteristics of candidate route 3 = 2 - (3 + 3 + 2) ÷ 3.

[0119] Optionally, the candidate navigation features of the three candidate routes recorded in Table 5 can better express the relative features of one candidate route and the remaining routes in the same candidate navigation data, relative to the candidate route data of the candidate routes.

[0120] Table 5

[0121]

[0122] Optionally, determining the candidate route characteristics of a candidate route based on candidate route data of at least one candidate route further includes: determining candidate road data in the candidate route data of at least one candidate route; wherein the candidate road data includes first road data and the total length of the candidate route; the first road data includes the segment length of the candidate road level and / or candidate road condition data; determining a first proportion of the candidate route based on the first road data and the total length of the candidate route; determining the average proportion of at least one candidate route based on the first proportion of the candidate route; and determining the candidate road characteristics of the candidate route based on the average proportion and the first proportion.

[0123] The candidate road data may include first road data and the total length of the candidate routes. The first road data may include the segment length of the candidate road level and / or candidate road condition data. The total length of the candidate routes is the distance data of the candidate routes.

[0124] Specifically, candidate road data included in the candidate route data is determined from the candidate route data of the candidate routes.

[0125] The first percentage can refer to the ratio of the first road data to the total length of the candidate routes.

[0126] Specifically, based on the first road data and the total length of the candidate routes, the ratio of the first road data to the total length of the candidate routes is determined as the first proportion of the first road data in the candidate road data of the candidate routes.

[0127] Here, the average percentage can refer to the average of the first percentage.

[0128] Specifically, based on the first proportion of at least one type of first road data included in the candidate road data of at least one candidate route, the first proportions of the same type of first road data in at least one candidate route are summed, and then the average value of the first proportions is obtained based on the number of candidate routes.

[0129] Candidate road features may include candidate road grade features and / or candidate road condition features.

[0130] Specifically, based on the average proportion of candidate routes and the first proportion, the difference between the first proportion of candidate routes and the average proportion is determined as the candidate road feature of the candidate route.

[0131] In another optional embodiment of the present invention, the first proportion of the candidate routes can be directly determined as the candidate road characteristics of the candidate routes. For example, as shown in Table 6, Table 6 records the candidate road level characteristics and candidate road condition characteristics of three candidate routes from a target starting point to a target ending point.

[0132] For example, taking the candidate route data recorded in Table 4 as an example, and using the first proportion of the candidate route as the candidate road feature, the candidate road level feature of candidate route 1 can include candidate expressway level feature and candidate national highway level feature. The candidate expressway level feature of candidate route 1 = 1252 ÷ 21718, the candidate expressway feature of candidate route 2 = 1058 ÷ 21313, and the candidate expressway feature of candidate route 3 = 1389 ÷ 26496.

[0133] Candidate national highway grade characteristics of candidate route 1 = 474 ÷ 21718, candidate passage grade characteristics of candidate route 2 = 546 ÷ 21313, candidate national highway grade of candidate route 3 = 758 ÷ 26496.

[0134] For example, combining the candidate route data recorded in Table 4, and determining the candidate road characteristics of the candidate routes based on the average proportion and the first proportion, the candidate road condition characteristics of candidate route 1 are as follows: 2011÷21718-[(2011÷21718)+(1856÷21313)+(1674÷26496)]÷3; the candidate road condition data of candidate route 2 are as follows: 1856÷21313-[(2011÷21718)+(1856÷21313)+(1674÷26496)]÷3; the candidate road condition characteristics of candidate route 3 are as follows: 1674÷26496-[(2011÷21718)+(1856÷21313)+(1674÷26496)]÷3.

[0135] Optionally, the candidate road features of the three candidate routes recorded in Table 6 can better express the relative features of one candidate route and the same candidate road data of the remaining routes compared to the candidate route data of the candidate routes.

[0136] Table 6

[0137] ID candidate highway grade features candidate national road grade features candidate road condition features 1 0.058 0.021 0.012 2 0.050 0.026 0.006 3 0.052 0.029 -0.018

[0138] S330. Train the route recommendation model based on the candidate route features and the ground truth penalty values ​​of the candidate routes.

[0139] The penalty score, in this context, refers to the similarity between the feasible routes provided by the navigation device from the user's starting point to the target destination and the actual routes taken by the user from the starting point to the target destination. In other words, it's data used to label the similarity relationship between candidate routes and actual routes. For example, if there are three candidate routes, the candidate route with the highest similarity to the actual route has a penalty score of 0, the candidate route with the second highest similarity has a penalty score of 1, and the candidate route with the lowest similarity has a penalty score of 2.

[0140] The route recommendation model can be a model pre-trained from an ensemble classifier; the ensemble classifier can refer to the XGBoosting algorithm. Furthermore, the parameters of the XGBoosting algorithm need to be reset. For example, the parameter settings for the XGBoosting algorithm can be as follows: Figure 1B As shown. The `booster` parameter refers to the model setting of the XGBoosting algorithm. In this embodiment, the XGBoosting model is preferably set to `gbtree`: a tree-based model. The `objective` parameter defines the loss function to be minimized. In this embodiment, at least one candidate path needs to be classified and predicted, so the `objective` parameter is preferably set to `multi:softmax`. The `num_class` parameter represents the number of class parameters. Preferably, in this embodiment, three candidate paths can be predicted, and `num_class` can be set to 3. The `max_depth` parameter refers to the maximum depth of the tree. Preferably, to prevent overfitting, `max_depth` can be set to 6 in this embodiment. The `eta` parameter is the learning rate of the value model. Preferably, to improve the stability of the trained model, `eta` can be set to 0.01 in this embodiment. The `gamma` parameter is a parameter used to optimize the model and reduce model redundancy. Preferably, in this embodiment, `gamma` can be set to 0.1. The `lambda` parameter is a regularization term parameter that controls the model complexity of the weight values. Preferably, in this embodiment, `lambda` can be set to 2.

[0141] It should be noted that during the training of the route recommendation model, a loss function can be preset. The training loss is calculated based on the predicted penalty value output by the route recommendation model using candidate route features and the ground truth penalty value of the candidate route. The training loss is then used to train the route recommendation model until a training stopping condition is met. The training stopping condition can be that the training loss stabilizes within a set range, or that the number of iterations meets a set requirement; the set range and the set number of iterations can be set by those skilled in the art based on the actual situation. It should be noted that the preset loss function can be the cross-entropy loss function.

[0142] Furthermore, to prevent the algorithm from overfitting during model training, which would lead to a decrease in the model's generalization ability, all training data (candidate route features) can be divided into two sets: one part of the data is used as the training set, and the other part is used as the test set. The division ratio can be 8:2, that is, 80% of the data is used for training, and 20% of the data is not used for training but is only used to test the accuracy of the training model.

[0143] In this invention, candidate navigation features and / or candidate route features are extracted from candidate path information of at least one candidate route and used as input to the route recommendation model to obtain the predicted penalty value of the candidate route. The route recommendation model is trained based on the predicted penalty value and the true penalty value of the candidate route. The above technical solution combines the candidate navigation features and candidate route features of the candidate route to determine the candidate route features from multiple dimensions, thereby improving the accuracy of the route recommendation model and thus improving the accuracy of route recommendation.

[0144] Based on the above embodiments, as an optional approach of the present invention, determining the penalty truth value of a candidate route may involve: obtaining the actual route between the target starting point and the target ending point; determining the similarity between the actual route and the candidate route based on the actual route and the candidate route; and determining the penalty truth value of the candidate route based on the similarity.

[0145] The actual route refers to the route the user actually took from the target origin to the target destination in their historical records. Optionally, the actual route taken by the user from the target origin to the target destination can be the same as or different from the candidate routes.

[0146] Specifically, the actual route taken by the user between the target starting point and the target ending point can be obtained. Then, for each candidate route, the similarity between the actual route and the candidate route can be determined, and the penalty truth value of the candidate route can be determined based on the similarity. For example, the penalty truth value of the candidate route can be determined based on the similarity and a preset standard. For example, if the similarity between the actual route and a certain candidate route is 0.98, then the penalty truth value of the candidate route can be determined to be 0.

[0147] Example 4

[0148] Figure 4 This is a schematic diagram of a route recommendation device provided in Embodiment 4 of the present invention. This embodiment is applicable to situations requiring accurate route recommendations to users. The route recommendation device can be implemented in hardware and / or software and can be integrated into an electronic device that carries the route recommendation function, such as a server. Figure 4 As shown, the device includes:

[0149] The candidate route determination module 410 is used to determine at least one candidate route from the target starting point to the target ending point.

[0150] The candidate route feature determination module 420 is used to determine the candidate route features of a candidate route based on the candidate route data of at least one candidate route. The candidate route features include candidate navigation features and / or candidate road features.

[0151] The target route recommendation module 430 is used to determine the target route from at least one candidate route based on the characteristics of the candidate routes using a route recommendation model, and recommend the target route to the user. The route recommendation model is a model trained on an ensemble classifier.

[0152] The technical solution of this invention, by determining the candidate route features of at least one candidate route and inputting the candidate route features of at least one candidate route into a route recommendation model, determines a target route from at least one candidate route and recommends it to the user. This solves the problem of unsatisfactory route recommendation results in the prior art, greatly improves the accuracy of route recommendation, and is more in line with the user's travel preferences.

[0153] Optionally, the target route recommendation module 430 is specifically used for:

[0154] Based on the route recommendation model, the route penalty score of the candidate route is determined according to the characteristics of the candidate route;

[0155] Based on the route penalty points, rank at least one candidate route;

[0156] Based on the sorting results, determine the target route.

[0157] Optionally, the candidate route feature determination module 420 is specifically used for:

[0158] Based on the candidate navigation data in the candidate route data of at least one candidate route, determine the mean of the candidate navigation data for at least one candidate route;

[0159] Based on the mean of the candidate navigation data and the candidate navigation data of the candidate routes, the candidate navigation features of the candidate routes are determined.

[0160] Optionally, the candidate route feature determination module 420 is also specifically used for:

[0161] Determine candidate road data from candidate route data for at least one candidate route; wherein, candidate road data includes first road data and total length of candidate route; first road data includes segment length of candidate road level and / or candidate road condition data;

[0162] Based on the first road data and the total length of the candidate routes, determine the first proportion of the candidate routes;

[0163] Based on the first percentage of candidate routes, determine the average percentage of at least one candidate route;

[0164] Based on the average percentage and the first percentage, the candidate road characteristics of the candidate routes are determined.

[0165] The route recommendation device provided in this embodiment of the invention can execute the route recommendation method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method execution.

[0166] Example 5

[0167] Figure 5 This is a schematic diagram of a training device for a route recommendation model provided in Embodiment 5 of the present invention. This embodiment is applicable to situations requiring accurate route recommendations to users. The training device for this route recommendation model can be implemented in hardware and / or software and can be integrated into an electronic device that carries the training function of the route recommendation model, such as a server. Figure 5 As shown, the device includes:

[0168] The candidate route determination module 510 is used to determine at least one candidate route from the target starting point to the target ending point.

[0169] The candidate route feature determination module 520 is used to determine the candidate route features of a candidate route based on the candidate route data of at least one candidate route. The candidate route features include candidate navigation features and / or candidate road features.

[0170] The route recommendation model training module 530 trains the route recommendation model based on the features of candidate routes and the ground truth penalty values ​​of candidate routes. The route recommendation model is a model obtained by pre-training an ensemble classifier.

[0171] In this invention, candidate navigation features and / or candidate route features are extracted from candidate path information of at least one candidate route and used as input to the route recommendation model to obtain the predicted penalty value of the candidate route. The route recommendation model is trained based on the predicted penalty value and the true penalty value of the candidate route. The above technical solution combines the candidate navigation features and candidate route features of the candidate route to determine the candidate route features from multiple dimensions, thereby improving the accuracy of the route recommendation model and thus improving the accuracy of route recommendation.

[0172] Optionally, the device further includes a penalty truth value determination module, which is used for:

[0173] Obtain the actual route between the target starting point and the target ending point;

[0174] Based on the actual route and the candidate routes, determine the similarity between the actual route and the candidate routes;

[0175] The ground truth penalty value for candidate routes is determined based on similarity.

[0176] The training apparatus for the route recommendation model provided in this embodiment of the invention can execute the training method for the route recommendation model provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0177] Example 6

[0178] Figure 6 A schematic diagram of an electronic device 10 that can be used to implement embodiments of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0179] like Figure 6 As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 12 or loaded from storage unit 18 into the RAM 13. The RAM 13 may also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0180] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0181] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as route recommendation methods or methods for training route recommendation models.

[0182] In some embodiments, the route recommendation method or route recommendation model training method may be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the route recommendation method or route recommendation model training method described above may be performed. Alternatively, in other embodiments, processor 11 may be configured to perform the route recommendation method or route recommendation model training method by any other suitable means (e.g., by means of firmware).

[0183] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0184] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0185] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0186] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0187] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or computing systems that include middleware components (e.g., application servers), or computing systems that include frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0188] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0189] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0190] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A route recommendation method, characterized in that, include: Identify at least one candidate route from the target starting point to the target ending point; Based on candidate route data from at least one candidate route, candidate route features are determined; wherein, the candidate route features include candidate navigation features and / or candidate road features; Based on the route recommendation model, a target route is determined from at least one candidate route according to the characteristics of the candidate routes, and the target route is recommended to the demander; wherein, the route recommendation model is a model obtained by pre-training an ensemble classifier; The step of determining the candidate route features based on candidate route data from at least one candidate route includes: Determine candidate road data from candidate route data for at least one candidate route; wherein, the candidate road data includes first road data and the total length of the candidate route; the first road data includes the segment length of the candidate road level and / or candidate road condition data; Based on the first road data and the total length of the candidate routes, a first proportion of the candidate routes is determined; Based on the first proportion of the candidate routes, determine the average proportion of at least one candidate route; Based on the average percentage and the first percentage, the candidate road features of the candidate routes are determined.

2. The method according to claim 1, characterized in that, Based on the route recommendation model, and according to the characteristics of the candidate routes, a target route is determined from at least one candidate route, including: Based on the route recommendation model, the route penalty score of the candidate route is determined according to the characteristics of the candidate route; Based on the route penalty points, at least one candidate route is ranked. Based on the sorting results, determine the target route.

3. The method according to claim 1, characterized in that, The step of determining the candidate route features based on candidate route data from at least one candidate route includes: Based on the candidate navigation data in the candidate route data of at least one candidate route, determine the mean of the candidate navigation data for at least one candidate route; Based on the mean of the candidate navigation data and the candidate navigation data of the candidate route, the candidate navigation features of the candidate route are determined.

4. A route recommendation device, characterized in that, include: The candidate route determination module is used to determine at least one candidate route from the target starting point to the target ending point; A candidate route feature determination module is used to determine candidate route features of the candidate route based on candidate route data of at least one candidate route; wherein, the candidate route features include candidate navigation features and / or candidate road features; The target route recommendation module is used to determine a target route from at least one candidate route based on the characteristics of the candidate routes using a route recommendation model, and recommend the target route to the demander; wherein, the route recommendation model is a model obtained by training an ensemble classifier; Specifically, the candidate route feature determination module is used for: Determine candidate road data from candidate route data for at least one candidate route; wherein, the candidate road data includes first road data and the total length of the candidate route; the first road data includes the segment length of the candidate road level and / or candidate road condition data; Based on the first road data and the total length of the candidate routes, a first proportion of the candidate routes is determined; Based on the first proportion of the candidate routes, determine the average proportion of at least one candidate route; Based on the average percentage and the first percentage, the candidate road features of the candidate routes are determined.

5. An electronic device, characterized in that, The electronic device includes: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the route recommendation method according to any one of claims 1-3.

6. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that cause a processor to execute the route recommendation method according to any one of claims 1-3.

Images