Travel scheme recommendation, model training method and device, equipment and storage medium

Abstract

The present disclosure provides a travel scheme recommendation method and device, a model training method and device, equipment and a storage medium, relates to the technical field of computers, and particularly relates to the technical field of intelligent transportation. The specific implementation scheme is as follows: obtaining a travel behavior feature determined according to a historical travel scheme of a target user; obtaining a scheme feature of a candidate travel scheme from a starting point to a destination of the target user; determining a recommendation degree of each candidate travel scheme according to the obtained travel behavior feature and the scheme feature; and recommending each candidate travel scheme to the target user according to the recommendation degree of each candidate travel scheme. The travel scheme recommendation method and device provided by the present disclosure can improve the accuracy of the recommendation.

Description

Technical Field

[0001] This disclosure relates to the field of computer technology, and more particularly to the field of intelligent transportation technology. Background Technology

[0002] As electronic map platforms evolve technologically, users wanting to travel from their origin to their destination can refer to various travel options provided by these platforms. However, if an electronic map platform offers users too many travel options, the decision-making cost for the user increases. To reduce this cost, current technologies typically reduce the number of travel options offered, for example, only providing the option with the lowest travel cost. Summary of the Invention

[0003] This disclosure provides a travel plan recommendation method, model training method, apparatus, device, and storage medium.

[0004] According to a first aspect of this disclosure, a method for recommending travel plans is provided, including:

[0005] Obtain travel behavior characteristics determined based on the target user's historical travel plans;

[0006] Obtain the characteristics of candidate travel routes for the target user from the origin to the destination;

[0007] Based on the obtained travel behavior characteristics and plan characteristics, the recommendation level of each candidate travel plan is determined;

[0008] Based on the recommendation rate of each candidate travel option, the candidate travel options are recommended to the target user.

[0009] According to a second aspect of this disclosure, a model training method is provided, comprising:

[0010] Obtain sample travel behavior characteristics determined based on the sample users' sample historical travel plans;

[0011] Determine the sample plan characteristics of the sample user's candidate travel plans from the sample departure point to the sample destination, and the target travel plan selected by the sample user among the sample candidate travel plans;

[0012] The obtained sample travel behavior features and sample plan features are input into a preset network model to obtain the recommendation degree of each sample candidate travel plan output by the network model.

[0013] Based on the numerical relationship between the recommendation scores of the target travel plan and the recommendation scores of other travel plans, a loss value is generated to train the network model, wherein the other travel plans are: plans other than the target travel plan among the sample candidate travel plans.

[0014] Based on the loss value, the model parameters of the network model are adjusted to obtain the recommendation rate acquisition model.

[0015] According to a third aspect of this disclosure, a travel plan recommendation device is provided, comprising:

[0016] The travel behavior feature acquisition module is used to obtain travel behavior features determined based on the target user's historical travel plans;

[0017] The plan feature acquisition module is used to obtain the plan features of the candidate travel plans of the target user from the origin to the destination;

[0018] The recommendation determination module is used to determine the recommendation level of each candidate travel plan based on the obtained travel behavior characteristics and plan characteristics.

[0019] The travel plan recommendation module is used to recommend each candidate travel plan to the target user based on the recommendation level of each candidate travel plan.

[0020] According to a fourth aspect of this disclosure, a model training apparatus is provided, comprising:

[0021] The sample travel behavior feature acquisition module is used to obtain sample travel behavior features determined based on the sample users' sample historical travel plans.

[0022] The feature and travel plan determination module is used to determine the sample plan features of the sample user's candidate travel plans from the sample departure point to the sample destination, and the target travel plan selected by the sample user among the sample candidate travel plans.

[0023] The recommendation degree acquisition module is used to input the obtained sample travel behavior features and sample plan features into a preset network model to obtain the recommendation degree of each sample candidate travel plan output by the network model.

[0024] The loss value generation module is used to generate a loss value for training the network model based on the numerical relationship between the recommendation degree of the target travel plan and the recommendation degree of other travel plans, wherein the other travel plans are: plans other than the target travel plan in the sample candidate travel plans;

[0025] The model acquisition module is used to adjust the model parameters of the network model based on the loss value to obtain the recommendation acquisition model.

[0026] According to a fifth aspect of this disclosure, an electronic device is provided, comprising:

[0027] At least one processor; and

[0028] A memory communicatively connected to the at least one processor; wherein,

[0029] The memory stores instructions that can be executed by the at least one processor, which enables the at least one processor to perform the aforementioned travel plan recommendation or model training methods.

[0030] According to a sixth aspect of this disclosure, a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to cause the computer to perform the aforementioned travel plan recommendation or model training methods.

[0031] According to a seventh aspect of this disclosure, a computer program product is provided, comprising a computer program that, when executed by a processor, implements the aforementioned travel plan recommendation or model training method.

[0032] As can be seen from the above, in the solution provided by this disclosure, the recommendation degree of the candidate travel plan is obtained based on the travel behavior characteristics determined by the historical travel plans and the plan characteristics of the candidate travel plan. In this way, the recommendation degree can reflect the relationship between the candidate travel plan and the historical travel plans. Since the historical travel plans can reflect the preferences shown by the user during the previous travel, the recommendation degree can represent the degree of conformity between the candidate travel plan and the user's preferences. The candidate travel plans that conform to the user's preferences are recommended to the user first according to the recommendation degree. The recommended candidate travel plans are more likely to be adopted by the user, thus improving the accuracy of the recommendation.

[0033] In addition, by recommending multiple candidate travel options to the target user and using the recommendation score as a reference, the target user can choose a travel option that better suits their needs based on the recommendation score, while ensuring that there are enough travel options available for making a final decision. This reduces the target user's decision-making cost and makes the decision-making process easier.

[0034] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0035] The accompanying drawings are provided to better understand this solution and do not constitute a limitation of this disclosure. Wherein:

[0036] Figure 1 This is a flowchart illustrating the first travel plan recommendation method provided in this embodiment of the disclosure;

[0037] Figure 2 This is a flowchart illustrating the second travel plan recommendation method provided in this embodiment of the disclosure;

[0038] Figure 3 This is a schematic flowchart of a model training method provided in an embodiment of this disclosure;

[0039] Figure 4a This is a flowchart of a method for obtaining model training samples provided in an embodiment of this disclosure;

[0040] Figure 4b This is a flowchart of a travel plan recommendation and model training method provided in an embodiment of this disclosure;

[0041] Figure 5 This is a schematic diagram of the structure of a travel plan recommendation device provided in an embodiment of this disclosure;

[0042] Figure 6 This is a schematic diagram of the structure of a model training device provided in an embodiment of this disclosure;

[0043] Figure 7 This is a block diagram of an electronic device used to implement the travel plan recommendation method or model training method of the embodiments of this disclosure. Detailed Implementation

[0044] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0045] The application scenarios of the solutions provided in the embodiments of this disclosure are described below.

[0046] The travel plan recommendation method provided in this disclosure can be applied to application software that includes electronic maps, such as electronic map platforms and navigation software.

[0047] In one scenario, a user can input their desired destination into navigation software installed on their electronic device. The navigation software then uses the device's positioning system to determine the user's current starting point. In this case, the navigation software can apply the travel plan recommendation method provided in this embodiment to offer the user multiple travel plans. After the user specifies a travel plan, the navigation software begins navigation according to that plan.

[0048] The travel plan recommendation method provided in this disclosure will be described in detail below through specific embodiments.

[0049] In one embodiment of this disclosure, see [link to embodiment]. Figure 1The document provides a flowchart of the first travel plan recommendation method, which includes the following steps S101-S104.

[0050] Step S101: Obtain travel behavior characteristics determined based on the target user's historical travel plans.

[0051] The aforementioned target users are the recommended users for travel plans.

[0052] Historical travel plans can be travel plans that the target user has used in the past period of time.

[0053] Travel behavior characteristics can be obtained based on the target user's interaction with historical travel plans. For example, the interaction could be the duration of the target user's browsing of various historical travel plans, or the mode of travel used in the final travel plan chosen by the target user.

[0054] Step S102: Obtain the characteristics of candidate travel routes for the target user from the origin to the destination.

[0055] Candidate travel plans are travel options to be recommended to the target user. One implementation can obtain candidate travel plans from the origin to the destination based on an existing route recommendation system. Another implementation can select historical travel plans with the origin and destination specified by the target user as candidate travel plans. Yet another implementation can select historical travel plans with the same origin and destination as the target user from all historical travel plans used by other users as candidate travel plans.

[0056] The characteristics of candidate travel plans can characterize their attributes and, more specifically, the costs a user incurs when adopting them, such as financial, time, and physical exertion. These attributes can include price, travel time, travel distance, mode of transportation used, segmented walking distances, and number of transfers.

[0057] Step S103: Determine the recommendation level of each candidate travel plan based on the obtained travel behavior characteristics and plan characteristics.

[0058] Recommendation score indicates how well each candidate travel option matches the target user. Recommendation score can be represented by a rating score or by an ordinal probability. The ordinal probability can be understood as the probability that a candidate travel option with this recommendation score will be recommended to the target user more frequently than other candidate travel options.

[0059] The following describes three ways to determine the recommendation level.

[0060] In the first implementation, the features of each candidate travel plan can be compared with the features of travel behavior to obtain the matching degree between each plan feature and each travel behavior feature. Then, the average matching degree is calculated using the obtained matching degree, and the recommendation degree of each candidate travel plan is set according to the average matching degree. The higher the average matching degree, the higher the recommendation degree of the candidate travel plan.

[0061] The matching degree can be calculated by the similarity between the feature vector representing the scheme features and the feature vector representing the travel behavior features.

[0062] In the second implementation, travel behavior features can be aggregated, and the recommendation level of candidate travel options can be determined based on the aggregated travel behavior sub-features and option features. For details, please refer to the following... Figure 2 The embodiments shown are not described in detail here.

[0063] In the third implementation method, in determining the recommendation degree of each candidate solution, in addition to considering the above-mentioned travel behavior characteristics and solution characteristics, additional characteristics of the target user can also be considered. For specific implementation methods, please refer to the subsequent embodiments, which will not be detailed here.

[0064] It should be noted that the embodiments disclosed herein are merely examples to illustrate the method of determining the degree of recommendation, and do not limit the method of determining the degree of recommendation.

[0065] Step S104: Recommend each candidate travel option to the target user based on the recommendation level of each candidate travel option.

[0066] In one embodiment of this disclosure, candidate travel plans can be sorted in descending order of recommendation level, and then recommended to the target user according to the sorting results. This allows candidate travel plans with higher recommendation levels to be displayed more prominently, making it easier for the target user to notice them. This also prioritizes recommending the top-ranked candidate travel plans to the target user.

[0067] As can be seen from the above, in the solution provided by this disclosure, the recommendation degree of the candidate travel plan is obtained based on the travel behavior characteristics determined by the historical travel plans and the plan characteristics of the candidate travel plan. In this way, the recommendation degree can reflect the relationship between the candidate travel plan and the historical travel plans. Since the historical travel plans can reflect the preferences shown by the user during the previous travel, the recommendation degree can represent the degree of conformity between the candidate travel plan and the user's preferences. The candidate travel plans that conform to the user's preferences are recommended to the user first according to the recommendation degree. The recommended candidate travel plans are more likely to be adopted by the user, thus improving the accuracy of the recommendation.

[0068] In addition, by recommending multiple candidate travel options to the target user and using the recommendation score as a reference, the target user can choose a travel option that better suits their needs based on the recommendation score, while ensuring that there are enough travel options available for making a final decision. This reduces the target user's decision-making cost and makes the decision-making process easier.

[0069] As described above, the recommendation level of candidate travel options can be determined through different methods. In one embodiment of this disclosure, see... Figure 2 Based on the aforementioned second method for determining the recommendation level, a second method for recommending travel plans is provided, which includes the following steps S201-S206.

[0070] Step S201: Obtain travel behavior characteristics determined based on the target user's historical travel plans.

[0071] Step S202: Obtain the characteristics of candidate travel routes for the target user from the origin to the destination.

[0072] The steps S201-S202 described above are the same as the steps S101-S102 described above, and will not be described in detail here.

[0073] Step S203: Aggregate the obtained travel behavior features according to the preset aggregation dimensions to obtain travel behavior sub-features.

[0074] Aggregating the obtained travel behavior features according to the aggregation dimension can be understood as follows: classifying the obtained travel behavior features according to the aggregation dimension to obtain different feature sets, each feature set including at least one travel behavior feature, and then aggregating the travel behavior features included in each feature set into a single feature, called a travel behavior sub-feature. Specifically, when aggregating the travel behavior features included in a feature set, feature concatenation can be used to combine the various travel behavior features end-to-end into a single feature. Alternatively, the average feature of the various travel behavior features can be calculated as the aggregated feature, or the sum of the various travel behavior features can be calculated as the aggregated feature.

[0075] The aggregation dimension can be either the time dimension or the spatial dimension of the travel plan. If the aggregation dimension is the time dimension, then the travel behavior features in each feature set are all taken from historical travel plans within the same time period, and the common attribute of the travel behavior features is the time period. If the aggregation dimension is the spatial dimension, then the travel behavior features in each feature set are all taken from historical travel plans within the same spatial region, and the common attribute of the travel behavior features is the spatial region, which can be represented as a geographical area defined on an electronic map.

[0076] Alternatively, both time and space dimensions can be used to aggregate travel behavior features, that is, to aggregate the behavioral features of candidate travel plans belonging to the same time period and the same spatial area to obtain travel behavior sub-features.

[0077] As can be seen from the above, the aggregation dimensions can be different. The travel behavior characteristics of target users can be aggregated according to the time dimension, so that the aggregated travel behavior sub-features can reflect the changes in users' travel preferences in different time periods. The travel behavior characteristics of users can also be aggregated according to the spatial dimension, so that the aggregated travel behavior sub-features can reflect the differences in the preferences of target users in different spatial regions. In the above cases, the travel behavior sub-features reflect the preferences of target users more in line with the actual preferences of target users under different conditions, and the obtained travel behavior sub-features are more accurate.

[0078] Alternatively, time and space dimensions can be used simultaneously for aggregation, so that the travel behavior sub-features can reflect the user preferences of target users when traveling in different time periods and different spatial areas, thereby further improving accuracy.

[0079] Step S204: For each candidate travel plan, calculate the matching degree between each travel behavior sub-feature and the candidate travel plan, and obtain the historical travel behavior features corresponding to the candidate travel plan based on the calculated matching degree and each travel behavior sub-feature.

[0080] The matching degree between travel behavior sub-features and candidate travel plans can be represented by the similarity between the features of the travel behavior sub-features and the features of the candidate travel plans. Specifically, feature vectors representing travel behavior sub-features and feature vectors representing candidate travel plan features can be obtained, and the similarity between the resulting feature vectors can be calculated as the matching degree. Specific calculation methods can include cosine similarity, Euclidean distance, etc., between feature vectors.

[0081] The following explains how historical travel behavior characteristics are obtained.

[0082] In the first implementation, the matching degree of each travel behavior sub-feature can be used as the weight to perform a weighted calculation on each travel behavior sub-feature, and the weighted calculation result can be used as the historical travel behavior feature corresponding to the candidate travel plan.

[0083] In this case, since the travel behavior sub-features with higher matching degree have larger weights, the weighted historical travel behavior features are quite similar to the travel behavior sub-features with higher matching degree. This allows the historical travel behavior features to have consistent representational capabilities with the travel behavior sub-features with higher matching degree. Furthermore, since the travel behavior sub-features can represent the target user's travel preferences reflected by travel behavior features in the same dimension under the aggregation dimension, the historical travel behavior features can also accurately represent the travel preferences that the target user may have.

[0084] In the second implementation, for each candidate travel plan, the matching degree between each travel behavior sub-feature and the candidate travel plan can be compared, and the travel behavior sub-feature with the highest matching degree can be used as the historical travel behavior feature of the candidate travel plan.

[0085] Step S205: Determine the recommendation level of each candidate travel plan based on the historical travel behavior characteristics and the plan characteristics of each candidate travel plan.

[0086] Since historical travel behavior characteristics can reflect the travel preferences of target users, and travel plans that match the travel preferences of target users are more likely to be ultimately chosen by them, candidate travel plans that match user travel preferences can be given a higher recommendation score.

[0087] To determine whether candidate travel plans match the travel preferences of target users, the similarity between historical travel behavior characteristics and plan characteristics can be calculated for each candidate travel plan. The higher the similarity, the more the candidate travel plan matches the travel preferences of target users reflected in historical travel behavior characteristics.

[0088] Step S206: Recommend each candidate travel option to the target user based on the recommendation level of each candidate travel option.

[0089] Step S206 is the same as step S104 described above, and will not be detailed here.

[0090] As can be seen from the above, in the solution provided by the embodiments of this disclosure, the historical travel behavior features corresponding to the candidate travel plan are obtained based on the calculated matching degree and each travel behavior sub-feature. In this way, the historical travel behavior features can reflect the travel preferences of the target user based on the travel behavior sub-features. When the recommendation degree of each candidate plan is determined by using the historical travel behavior features and plan features, the recommendation degree is related to the travel preferences of the target user, making the determined recommendation degree more meaningful for the target user.

[0091] The third method for determining the recommendation level in step S103 above will be explained below.

[0092] In this implementation, additional characteristics of the target user can be obtained, wherein the additional characteristics may include at least one of the following characteristics:

[0093] The first type of feature is the scene characteristics of the environment in which the target user makes a travel request from the origin to the destination. For example, the weather characteristics and location characteristics of the scene in which the target user makes the travel request.

[0094] The second type of feature: target user characteristics. These user characteristics can be obtained based on the target user's attributes, such as gender and age.

[0095] The third type of feature: preference features reflecting the travel preferences of the target users. These preference features can be obtained by counting the number or proportion of times users search for different modes of transportation. For example, the preference features can be obtained by statistically analyzing the proportion of times users search for ride-hailing on electronic map platforms out of the total number of times they search for different modes of transportation.

[0096] The fourth characteristic is the proportion of different types of travel options among those recommended to the target user. This characteristic can be obtained by counting the number of times each type of option is recommended to the target user. The type of option can be determined based on the travel modes included in the travel option. The travel option can include a single mode of travel or a combination of modes of travel, such as walking followed by public transportation.

[0097] The fifth feature: This characterizes the position of the travel option selected by the target user within the recommended travel options. For example, if there is an existing ranking sequence of travel options, the position feature can be determined based on the ranking order of the travel option selected by the target user.

[0098] If at least one of the above additional features is obtained, the recommendation level of each candidate travel plan can be determined based on the obtained travel behavior features, plan features, and additional features.

[0099] As can be seen from the above, in the solution provided by the embodiments of this disclosure, additional features are used when determining the recommendation degree of each candidate travel plan. This makes the recommendation degree affected by multiple aspects of information such as target user information, travel plan information, and travel scenario information represented by the additional features. This makes the determined recommendation degree more consistent with the actual travel situation of the target user, and the travel plan recommendation based on the recommendation degree is more accurate.

[0100] In one embodiment of this disclosure, the aforementioned step S103, which determines the recommendation degree of each candidate travel plan based on the obtained travel behavior characteristics and plan characteristics, can be implemented based on a network model, as follows:

[0101] The obtained travel behavior features and plan features are input into a pre-trained recommendation model to obtain the recommendation degree of each candidate travel plan output by the recommendation degree model. The recommendation degree model is a model used to obtain the recommendation degree of travel plans by training a preset network model with the travel behavior features and plan features of sample users. The plan features are the plan features of sample candidate travel plans from the sample departure point to the sample destination.

[0102] The method for obtaining the travel behavior characteristics of sample users is similar to that for obtaining the travel behavior characteristics of target users in step S101 above. The only difference is the substitution of the names of sample users and target users, which will not be described in detail here.

[0103] The specific training method for training the above recommendation model can be found in the subsequent embodiments, and will not be detailed here.

[0104] As can be seen from the above, in this embodiment of the disclosure, the recommendation degree of each candidate travel plan is obtained by a pre-trained recommendation degree acquisition model. The recommendation degree acquisition model is trained using the travel behavior features of sample users and the features of sample plans, and has learned the data processing capability to obtain the recommendation degree. Therefore, the recommendation degree obtained by using the recommendation degree acquisition model is relatively accurate.

[0105] The specific training method for obtaining the recommendation score model is explained below.

[0106] In one embodiment of this disclosure, see [link to embodiment]. Figure 3 A flowchart of a model training method is provided, which includes the following steps S301-S305.

[0107] Step S301: Obtain the sample travel behavior characteristics determined based on the sample users' sample historical travel plans.

[0108] Step S302: Determine the sample plan characteristics of the sample candidate travel plans from the sample origin to the sample destination, and the target travel plan selected by the sample user among the sample candidate travel plans.

[0109] Steps S301-S302 are similar to steps S101-S102, except for the substitution of names and concepts such as sample user and target user, sample historical travel plan and historical travel plan, sample travel behavior characteristics and travel behavior characteristics, sample plan characteristics and plan characteristics, which will not be described in detail here.

[0110] Step S303: Input the obtained sample travel behavior features and sample plan features into the preset network model to obtain the recommendation degree of each sample candidate travel plan output by the network model.

[0111] The feature vectors representing the travel behavior characteristics of the samples and the feature vectors representing the travel plans of the samples are input as parameters into the network model. The network model processes the data and outputs the recommendation degree of the sample candidate travel plans.

[0112] The data processing procedure for the network model can be found in subsequent embodiments, and will not be detailed here.

[0113] Step S304: Generate a loss value for training the network model based on the numerical relationship between the recommendation scores of the target travel plan and the recommendation scores of other travel plans.

[0114] Other travel options are those other than the target travel option among the sample candidate travel options.

[0115] The numerical relationship indicates the priority of the target travel option compared to other travel options when recommended to the user; the higher the recommendation level, the higher the priority of the travel option.

[0116] For example, when the recommendation level represents the order probability, a pairwise method can be used to compare each candidate travel option pairwise, with the numerical relationship representing the order probability. In this case, the recommendation order of each candidate option with other candidate options is determined, and a ranking sequence of candidate travel options is formed based on the resulting priority order.

[0117] The loss value can be determined based on the difference between the actual user's selected travel plan and the recommendation level of that travel plan. Specifically, the loss value can be obtained by calculating the cross-entropy loss as described above.

[0118] Step S305: Based on the loss value, adjust the model parameters of the network model to obtain the recommendation score model.

[0119] Adjusting the model parameters of a network model based on the loss value can be achieved by determining the parameter values ​​of the network model that minimize the loss value as the parameter values ​​of the model parameters of the model that obtains the recommendation score after training. Alternatively, an upper limit can be preset for the number of times the model parameters can be adjusted. Once the upper limit is reached, the parameter values ​​of the model parameters will be determined as the parameter values ​​of the model parameters of the model that obtains the recommendation score after training.

[0120] A network model whose parameters have been adjusted using the above method can be used as a model for obtaining recommendation scores.

[0121] As can be seen from the above, in the solution provided by the embodiments of this disclosure, a loss value is generated based on the numerical relationship between the recommendation degree of the target travel plan and the recommendation degree of other travel plans, and the model parameters of the network model are adjusted according to the loss value, so that the loss of the recommendation degree output by the adjusted model is smaller, that is, the output recommendation degree is more in line with the expected numerical relationship between the recommendation degrees of each candidate travel plan, thereby improving the accuracy of the obtained recommendation degree.

[0122] The following describes the data processing process of the network model after the obtained sample travel behavior features and sample plan features are input into the preset network model in the aforementioned step S303.

[0123] In one embodiment of this disclosure, the network model processes the data in the following manner and outputs the recommendation score for each sample candidate travel plan:

[0124] The obtained sample travel behavior features are aggregated according to a preset aggregation dimension to obtain sample travel behavior sub-features. For each sample candidate travel plan, the matching degree between each sample travel behavior sub-feature and the sample candidate travel plan is calculated. Based on the calculated matching degree and each sample travel behavior sub-feature, the sample historical travel behavior features corresponding to the sample candidate travel plan are obtained. Based on the sample historical travel behavior features corresponding to each sample candidate travel plan and the sample plan features of each sample candidate travel plan, the recommendation degree of each sample candidate travel plan is determined.

[0125] The data processing process of the network model is similar to the aforementioned steps S203-S205, with the only difference being the replacement of names and concepts such as sample travel behavior features and travel behavior characteristics, sample travel behavior sub-features and travel behavior sub-features, sample candidate travel schemes and candidate travel schemes, and sample historical travel behavior features and historical travel behavior features. These will not be described in detail here.

[0126] As can be seen from the above, in the scheme provided by the embodiments of this disclosure, the recommendation degree of each sample candidate travel scheme is determined according to the sample historical travel behavior characteristics corresponding to each sample candidate travel scheme and the sample scheme characteristics of each sample candidate travel scheme. The recommendation degree can represent the degree of conformity between each sample candidate scheme and the travel preferences reflected by the sample historical travel scheme of the target user. The network model trained according to this scheme can learn the data processing ability to accurately obtain the recommendation degree.

[0127] In another embodiment of this disclosure, during the data processing of the above-mentioned network model, based on the calculated matching degree and the travel behavior sub-features of each sample, the sample's historical travel behavior features corresponding to the candidate travel scheme are obtained, including:

[0128] Using the matching degree of each sample's travel behavior sub-features as weight, a weighted calculation is performed on each sample's travel behavior sub-features, and the weighted calculation result is used as the sample's historical travel behavior feature corresponding to the candidate travel plan of that sample.

[0129] The method of obtaining the sample's historical travel behavior features is similar to the first implementation method in step S204 above. The only difference is the replacement of the names of the sample travel behavior sub-features and travel behavior sub-features, sample historical travel behavior features and historical travel behavior features, etc., which will not be described in detail here.

[0130] As can be seen from the above, in the solution provided by the embodiments of this disclosure, by weighting, the weighted sample historical travel behavior features are more similar to the sample travel behavior sub-features with high matching degree, so that the sample historical travel behavior features can have consistent representational ability with the sample travel behavior sub-features with high matching degree. Since the sample travel behavior sub-features are obtained by aggregating the sample travel behavior features, they can reflect the travel preferences of the sample users. Therefore, the sample historical travel behavior features can more accurately represent the travel preferences that users may have.

[0131] In another embodiment of this disclosure, similar to step S103, when determining the recommendation level of the sample candidate travel plans, additional features of the sample users can be used, as follows:

[0132] Obtain additional sample features of the sample users, wherein the additional sample features include at least one of the following features:

[0133] Scene characteristics of the scenario in which sample users generate travel requests from the sample's origin to the sample's destination;

[0134] User characteristics of the sample users;

[0135] Preference characteristics that reflect the travel preferences of sample users;

[0136] The proportion of different types of travel options that have been recommended to sample users;

[0137] The location features of the selected travel options by the sample user appearing in the recommended travel options.

[0138] The additional characteristics of the sample users are similar to those of the target users in the previous embodiments, the only difference being the substitution of names such as sample users and target users, which will not be described in detail here.

[0139] The obtained sample travel behavior features and sample route features are input into a pre-defined network model to obtain the recommendation scores of each sample candidate travel route output by the network model, including:

[0140] The obtained sample travel behavior features, sample plan features, and additional features are input into a pre-defined network model to obtain the recommendation degree of each sample candidate travel plan output by the network model.

[0141] As can be seen from the above, in the solution provided by the embodiments of this disclosure, additional features are used when determining the recommendation degree of each sample candidate travel plan. This makes the recommendation degree affected by various aspects of information such as sample user information, travel plan information, and travel scenario information represented by the additional features, so that the determined recommendation degree is more in line with the actual travel situation of the sample users and the obtained recommendation degree is more accurate.

[0142] The following is through Figure 4a , Figure 4b The diagram shown illustrates the overall process of travel plan recommendation and model training methods based on a network model.

[0143] Figure 4a In this context, the obtained model training samples serve as the input to the network model during training, namely, the samples formed from the sample travel behavior features, sample scheme features, and additional features of the sample users in the aforementioned embodiments. The sources of each type of feature are shown above the model training samples.

[0144] The user behavior sample collection module is used to obtain short-term behavioral history of the scenario, namely the sample users' historical travel plans in step S301 above, and user travel selection preferences, namely the sample users' travel preferences. The sample travel behavior features determined by the sample historical travel plans are aggregated according to spatiotemporal behavior, that is, in the above embodiment, they are aggregated simultaneously according to the time dimension and the spatial dimension to obtain sample travel behavior sub-features; the sample users' travel preferences can determine preference features.

[0145] The user attributes in the figure are used to obtain the user characteristics of the sample users in the aforementioned embodiments.

[0146] The request scenario context in the diagram represents the scenario features of the scenario in which a sample user generates a travel request from the sample's origin to its destination.

[0147] The scheme attributes in the figure are used to obtain the proportion of different types of schemes among the travel schemes recommended to the sample users, and to characterize the position of the travel scheme selected by the sample user in the recommended travel schemes.

[0148] Figure 4b In this model training sample, the spatiotemporal behavior features are sample travel behavior sub-features obtained by aggregating spatiotemporal behavior. Behavior modeling is used to obtain sample historical travel behavior features according to the method of the aforementioned embodiment. Other features are other parts of the model training sample besides the sample travel behavior sub-features.

[0149] By inputting the feature vectors of historical travel behavior characteristics and other feature vectors into the model, model training can be performed. Figure 3 The solution provided in the example yields a recommendation rate acquisition model.

[0150] The online prediction process involves determining the recommendation level of each candidate travel option when a user submits a travel request, as described in S101-S103. The score used to rank the options is one form of recommendation level. This module ranks the candidate travel options according to their recommendation level and recommends each option to the target user.

[0151] Corresponding to the above-described travel plan recommendation method, this disclosure also recommends a travel plan recommendation device.

[0152] In one embodiment of this disclosure, see [link to embodiment]. Figure 5 A schematic diagram of a travel plan recommendation device is provided, comprising:

[0153] The travel behavior feature acquisition module 501 is used to obtain travel behavior features determined based on the target user's historical travel plans;

[0154] The scheme feature acquisition module 502 is used to obtain the scheme features of the candidate travel schemes of the target user from the origin to the destination;

[0155] The recommendation degree determination module 503 is used to determine the recommendation degree of each candidate travel plan based on the obtained travel behavior characteristics and plan characteristics;

[0156] The travel plan recommendation module 504 is used to recommend each candidate travel plan to the target user based on the recommendation degree of each candidate travel plan.

[0157] As can be seen from the above, in the solution provided by this disclosure, the recommendation degree of the candidate travel plan is obtained based on the travel behavior characteristics determined by the historical travel plans and the plan characteristics of the candidate travel plan. In this way, the recommendation degree can reflect the relationship between the candidate travel plan and the historical travel plans. Since the historical travel plans can reflect the preferences shown by the user during the previous travel, the recommendation degree can represent the degree of conformity between the candidate travel plan and the user's preferences. The candidate travel plans that conform to the user's preferences are recommended to the user first according to the recommendation degree. The recommended candidate travel plans are more likely to be adopted by the user, thus improving the accuracy of the recommendation.

[0158] In addition, by recommending multiple candidate travel options to the target user and using the recommendation score as a reference, the target user can choose a travel option that better suits their needs based on the recommendation score, while ensuring that there are enough travel options available for making a final decision. This reduces the target user's decision-making cost and makes the decision-making process easier.

[0159] In one embodiment of this disclosure, the recommendation determination module 503 includes:

[0160] The sub-feature acquisition unit is used to aggregate the obtained travel behavior features according to a preset aggregation dimension to obtain travel behavior sub-features.

[0161] The historical travel behavior feature acquisition unit is used to calculate the matching degree between each travel behavior sub-feature and the candidate travel plan for each candidate travel plan, and obtain the historical travel behavior features corresponding to the candidate travel plan based on the calculated matching degree and each travel behavior sub-feature.

[0162] The recommendation degree determination unit is used to determine the recommendation degree of each candidate travel plan based on the historical travel behavior characteristics and the plan characteristics of each candidate travel plan.

[0163] As can be seen from the above, in the solution provided by the embodiments of this disclosure, the historical travel behavior features corresponding to the candidate travel plan are obtained based on the calculated matching degree and each travel behavior sub-feature. This allows the historical travel behavior features to reflect the travel preferences of the target user based on the travel behavior sub-features. When the recommendation degree of each candidate plan is determined using the historical travel behavior features and plan features, the recommendation degree is related to the travel preferences of the target user, making the determined recommendation degree more meaningful for the target user.

[0164] In one embodiment of this disclosure, the historical travel behavior feature acquisition unit is specifically used to calculate the matching degree between each travel behavior sub-feature and the candidate travel plan for each candidate travel plan, and to perform weighted calculation on each travel behavior sub-feature using the matching degree of each travel behavior sub-feature as the weight, and to use the weighted calculation result as the historical travel behavior feature corresponding to the candidate travel plan.

[0165] In this case, since the travel behavior sub-features with higher matching degree have larger weights, the weighted historical travel behavior features are quite similar to the travel behavior sub-features with higher matching degree. This allows the historical travel behavior features to have consistent representational capabilities with the travel behavior sub-features with higher matching degree. Furthermore, since the travel behavior sub-features can represent the target user's travel preferences reflected by travel behavior features in the same dimension under the aggregation dimension, the historical travel behavior features can also accurately represent the travel preferences that the target user may have.

[0166] In one embodiment of this disclosure, the aggregation dimension includes: the time dimension of the travel plan and / or the regional dimension of the travel plan.

[0167] As can be seen from the above, the aggregation dimensions can be different. The travel behavior characteristics of target users can be aggregated according to the time dimension, so that the aggregated travel behavior sub-features can reflect the changes in users' travel preferences in different time periods. The travel behavior characteristics of users can also be aggregated according to the spatial dimension, so that the aggregated travel behavior sub-features can reflect the differences in the preferences of target users in different spatial regions. In the above cases, the travel behavior sub-features reflect the preferences of target users more in line with the actual preferences of target users under different conditions, and the obtained travel behavior sub-features are more accurate.

[0168] Alternatively, time and space dimensions can be used simultaneously for aggregation, so that the travel behavior sub-features can reflect the user preferences of target users when traveling in different time periods and different spatial areas, thereby further improving accuracy.

[0169] In one embodiment of this disclosure, the apparatus further includes:

[0170] The attachment feature acquisition module is used to acquire additional features of the target user, wherein the additional features include at least one of the following features:

[0171] The scene characteristics of the scenario in which the target user is located when the user makes a travel request from the departure point to the destination;

[0172] The user characteristics of the target user;

[0173] Preference characteristics that reflect the travel preferences of the target users;

[0174] The proportion of different types of travel options that have been recommended to the target user;

[0175] Location features that characterize the location of the travel plan selected by the target user among the recommended travel plans;

[0176] The recommendation degree determination module 503 is specifically used to determine the recommendation degree of each candidate travel plan based on the obtained travel behavior characteristics, plan characteristics and additional characteristics.

[0177] As can be seen from the above, in the solution provided by the embodiments of this disclosure, additional features are used when determining the recommendation degree of each candidate travel plan. This makes the recommendation degree affected by multiple aspects of information such as target user information, travel plan information, and travel scenario information represented by the additional features. This makes the determined recommendation degree more consistent with the actual travel situation of the target user, and the travel plan recommendation based on the recommendation degree is more accurate.

[0178] In one embodiment of this disclosure, the recommendation degree determination module 503 is specifically used to input the obtained travel behavior features and plan features into a pre-trained recommendation degree acquisition model to obtain the recommendation degree of each candidate travel plan output by the recommendation degree acquisition model. The recommendation degree acquisition model is a model for obtaining the recommendation degree of travel plans by training a preset network model using the travel behavior features and sample plan features of sample users. The sample plan features are the plan features of the sample candidate travel plans of the sample users from the sample departure point to the sample destination.

[0179] As can be seen from the above, in this embodiment of the disclosure, the recommendation degree of each candidate travel plan is obtained by a pre-trained recommendation degree acquisition model. The recommendation degree acquisition model is trained using the travel behavior features of sample users and the features of sample plans, and has learned the data processing capability to obtain the recommendation degree. Therefore, the recommendation degree obtained by using the recommendation degree acquisition model is relatively accurate.

[0180] Corresponding to the above-described model training method, this disclosure also provides a model training apparatus.

[0181] In one embodiment of this disclosure, see [link to embodiment]. Figure 6 A schematic diagram of a model training device is provided, comprising:

[0182] The sample travel behavior feature acquisition module 601 is used to obtain sample travel behavior features determined based on the sample users' sample historical travel plans.

[0183] The feature and travel plan determination module 602 is used to determine the sample plan features of the sample candidate travel plans of the sample user from the sample departure point to the sample destination, and the target travel plan selected by the sample user in the sample candidate travel plans.

[0184] The recommendation degree acquisition module 603 is used to input the obtained sample travel behavior features and sample plan features into a preset network model to obtain the recommendation degree of each sample candidate travel plan output by the network model.

[0185] The loss value generation module 604 is used to generate a loss value for training the network model based on the numerical relationship between the recommendation degree of the target travel plan and the recommendation degree of other travel plans, wherein the other travel plans are: plans other than the target travel plan in the sample candidate travel plans;

[0186] The model acquisition module 605 is used to adjust the model parameters of the network model based on the loss value to obtain the recommendation acquisition model.

[0187] As can be seen from the above, in the solution provided by the embodiments of this disclosure, a loss value is generated based on the numerical relationship between the recommendation degree of the target travel plan and the recommendation degree of other travel plans, and the model parameters of the network model are adjusted according to the loss value, so that the loss of the recommendation degree output by the adjusted model is smaller, that is, the output recommendation degree is more in line with the expected numerical relationship between the recommendation degrees of each candidate travel plan, thereby improving the accuracy of the obtained recommendation degree.

[0188] In one embodiment of this disclosure, the recommendation degree acquisition module 603 is specifically used to input the obtained sample travel behavior features and sample plan features into a preset network model, whereby the network model processes the data and outputs the recommendation degree of each sample candidate travel plan. The network model includes:

[0189] The sample travel behavior sub-feature acquisition module is used to aggregate the obtained sample travel behavior features according to a preset aggregation dimension to obtain sample travel behavior sub-features.

[0190] The sample historical travel behavior feature acquisition module is used to calculate the matching degree between each sample travel behavior sub-feature and the sample candidate travel plan for each sample candidate travel plan, and obtain the sample historical travel behavior features corresponding to the sample candidate travel plan based on the calculated matching degree and each sample travel behavior sub-feature.

[0191] The sample recommendation degree determination module is used to determine the recommendation degree of each sample candidate travel plan based on the sample historical travel behavior characteristics corresponding to each sample candidate travel plan and the sample plan characteristics of each sample candidate travel plan.

[0192] As can be seen from the above, in the scheme provided by the embodiments of this disclosure, the recommendation degree of each sample candidate travel scheme is determined according to the sample historical travel behavior characteristics corresponding to each sample candidate travel scheme and the sample scheme characteristics of each sample candidate travel scheme. The recommendation degree can represent the degree of conformity between each sample candidate scheme and the travel preferences reflected by the sample historical travel scheme of the target user. The network model trained according to this scheme can learn the data processing ability to accurately obtain the recommendation degree.

[0193] In one embodiment of this disclosure, the sample historical travel behavior feature acquisition module is specifically used to calculate the matching degree between each sample travel behavior sub-feature and the sample candidate travel plan for each sample candidate travel plan, and to perform weighted calculation on each sample travel behavior sub-feature using the matching degree of each sample travel behavior sub-feature as the weight, and to use the weighted calculation result as the sample historical travel behavior feature corresponding to the sample candidate travel plan.

[0194] As can be seen from the above, in the solution provided by the embodiments of this disclosure, by weighting, the weighted sample historical travel behavior features are more similar to the sample travel behavior sub-features with high matching degree, so that the sample historical travel behavior features can have consistent representational ability with the sample travel behavior sub-features with high matching degree. Since the sample travel behavior sub-features are obtained by aggregating the sample travel behavior features, they can reflect the travel preferences of the sample users. Therefore, the sample historical travel behavior features can more accurately represent the travel preferences that users may have.

[0195] In one embodiment of this disclosure, the apparatus further includes:

[0196] An additional feature acquisition module is used to obtain sample additional features of the sample users, wherein the sample additional features include at least one of the following features:

[0197] The scene characteristics of the scenario in which the sample user generates a travel request from the sample's origin to the sample's destination;

[0198] User characteristics of the sample users;

[0199] Preference features that reflect the travel preferences of the sample users;

[0200] The proportion of different types of travel options that have been recommended to the sample users;

[0201] The location feature that represents the position of the travel plan selected by the sample user in the recommended travel plans;

[0202] The recommendation degree acquisition module 603 is specifically used to input the obtained sample travel behavior features, sample plan features and additional features into a preset network model to obtain the recommendation degree of each sample candidate travel plan output by the network model.

[0203] As can be seen from the above, in the solution provided by the embodiments of this disclosure, additional features are used when determining the recommendation degree of each sample candidate travel plan. This makes the recommendation degree affected by various aspects of information such as sample user information, travel plan information, and travel scenario information represented by the additional features, so that the determined recommendation degree is more in line with the actual travel situation of the sample users and the obtained recommendation degree is more accurate.

[0204] The collection, storage, use, processing, transmission, provision, and disclosure of the target user's personal information in this technical solution comply with relevant laws and regulations and do not violate public order and good morals.

[0205] According to embodiments of this disclosure, this disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0206] In one embodiment of this disclosure, an electronic device is provided, comprising:

[0207] At least one processor; and

[0208] A memory communicatively connected to the at least one processor; wherein,

[0209] The memory stores instructions that can be executed by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform the travel plan recommendation or model training methods described in the foregoing method embodiments.

[0210] In one embodiment of this disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to cause the computer to perform the travel plan recommendation or model training methods described in the foregoing method embodiments.

[0211] In one embodiment of this disclosure, a computer program product is provided, including a computer program that, when executed by a processor, implements the travel plan recommendation or model training method described in the foregoing method embodiments.

[0212] Figure 7 A schematic block diagram of an example electronic device 700 that can be used to implement embodiments of the present disclosure 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 may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, 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 present disclosure described and / or claimed herein.

[0213] like Figure 7 As shown, device 700 includes a computing unit 701, which can perform various appropriate actions and processes based on a computer program stored in read-only memory (ROM) 702 or a computer program loaded from storage unit 708 into random access memory (RAM) 703. RAM 703 may also store various programs and data required for the operation of device 700. The computing unit 701, ROM 702, and RAM 703 are interconnected via bus 704. Input / output (I / O) interface 705 is also connected to bus 704.

[0214] Multiple components in device 700 are connected to I / O interface 705, including: input unit 706, such as keyboard, mouse, etc.; output unit 707, such as various types of monitors, speakers, etc.; storage unit 708, such as disk, optical disk, etc.; and communication unit 709, such as network card, modem, wireless transceiver, etc. Communication unit 709 allows device 700 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0215] The computing unit 701 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 701 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 computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 701 performs the various methods and processes described above, such as travel plan recommendation methods or model training methods. For example, in some embodiments, the travel plan recommendation method or model training method can be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 708. In some embodiments, part or all of the computer program can be loaded and / or installed on device 700 via ROM 702 and / or communication unit 709. When the computer program is loaded into RAM 703 and executed by the computing unit 701, one or more steps of the travel plan recommendation method or model training method described above can be performed. Alternatively, in other embodiments, the computing unit 701 may be configured, by any other suitable means (e.g., by means of firmware), to perform a travel plan recommendation method or a model training method.

[0216] 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), complex 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.

[0217] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0218] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of 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 fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0219] To provide interaction with a target user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the target user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the target user provides input to the computer. Other types of devices can also be used to provide interaction with the target user; for example, feedback provided to the target user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the target user can be received in any form (including sound input, voice input, or tactile input).

[0220] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a target user computer with a graphical user interface or a web browser through which the target user can interact with the 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., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0221] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact via communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other. Servers can be cloud servers, servers in distributed systems, or servers incorporating blockchain technology.

[0222] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0223] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A travel scheme recommendation method, comprising: obtaining travel behavior features determined according to historical travel schemes of a target user, the travel behavior features being obtained according to interaction behaviors of the target user with respect to the historical travel schemes, the interaction behaviors including: a time length for which the target user browses each historical travel scheme, and a travel mode adopted in a travel scheme finally selected by the target user; obtaining scheme features of candidate travel schemes from a departure location to a destination of the target user, the scheme features of the candidate travel schemes representing attributes of the candidate travel schemes, the attributes of the candidate travel schemes including at least one of the following attributes: a price, a travel time, a travel distance, a travel mode adopted in the candidate travel scheme, a segmented walking distance, and a number of transfers; aggregating the obtained travel behavior features according to a preset aggregation dimension to obtain travel behavior sub-features; for each candidate travel scheme, calculating a matching degree of each travel behavior sub-feature with respect to the candidate travel scheme, and obtaining historical travel behavior features corresponding to the candidate travel scheme based on the calculated matching degrees and the travel behavior sub-features; determining recommendation degrees of the candidate travel schemes according to the historical travel behavior features corresponding to the candidate travel schemes and the scheme features of the candidate travel schemes; and recommending the candidate travel schemes to the target user according to the recommendation degrees of the candidate travel schemes.

2. The method of claim 1, wherein, The obtaining of the historical travel behavior features corresponding to the candidate travel scheme based on the calculated matching degrees and the travel behavior sub-features comprises: performing weighted calculation on the travel behavior sub-features with the matching degrees of the travel behavior sub-features as weights, and taking a result of the weighted calculation as the historical travel behavior features corresponding to the candidate travel scheme.

3. The method of claim 1, wherein the aggregation dimension includes a time dimension of a travel scheme and / or a region dimension of a travel scheme.

4. The method of any one of claims 1-3, further comprising: obtaining additional features of the target user, wherein the additional features include at least one of the following features: a scene feature of a scene in which the target user is located when the target user generates a travel request from the departure location to the destination; a user feature of the target user; a preference feature reflecting a travel preference of the target user; a proportion feature of different types of travel schemes that have been recommended to the target user; a position feature representing a position of a travel scheme selected by the target user in recommended travel schemes; The determining of the recommendation degrees of the candidate travel schemes according to the obtained travel behavior features and the scheme features comprises: determining the recommendation degrees of the candidate travel schemes according to the obtained travel behavior features, the scheme features, and the additional features.

5. The method of any one of claims 1-3, wherein, The determining of the recommendation degrees of the candidate travel schemes according to the obtained travel behavior features and the scheme features comprises: input the obtained travel behavior features and the scheme features into a pre-trained recommendation degree obtaining model to obtain a recommendation degree of each candidate travel scheme output by the recommendation degree obtaining model, wherein the recommendation degree obtaining model is a model for obtaining a travel scheme recommendation degree obtained by training a preset network model using travel behavior features and sample scheme features of sample users, and the sample scheme features are scheme features of sample candidate travel schemes of the sample users from sample starting locations to sample destinations.

6. A model training method, comprising: obtaining sample travel behavior features determined according to sample historical travel schemes of sample users, the sample travel behavior features being obtained according to interaction behaviors of the sample users with respect to the sample historical travel schemes, the interaction behaviors including a time length for which the sample users browse each sample historical travel scheme and a travel mode adopted in a travel scheme finally selected by the sample users; determining sample scheme features of sample candidate travel schemes of the sample users from sample starting locations to sample destinations and a target travel scheme selected by the sample users in the sample candidate travel schemes, the scheme features of the candidate travel schemes representing attributes of the candidate travel schemes, the attributes including at least one of the following attributes: a price, a travel time, a travel distance, a travel mode adopted in the candidate travel scheme, a segmented walking distance, and a number of transfers; after inputting the obtained sample travel behavior features and the sample scheme features into a preset network model, performing data processing on the network model according to the following manner and outputting a recommendation degree of each sample candidate travel scheme: aggregating the obtained sample travel behavior features according to a preset aggregation dimension to obtain sample travel behavior sub-features; for each sample candidate travel scheme, calculating a matching degree of each sample travel behavior sub-feature with the sample candidate travel scheme, and obtaining sample historical travel behavior features corresponding to the sample candidate travel scheme based on the calculated matching degrees and the sample travel behavior sub-features; determining a recommendation degree of each sample candidate travel scheme according to the sample historical travel behavior features corresponding to the sample candidate travel schemes and the sample scheme features of the sample candidate travel schemes; generating a loss value for training the network model according to a numerical relationship between the recommendation degree of the target travel scheme and recommendation degrees of other travel schemes, wherein the other travel schemes are schemes other than the target travel scheme in the sample candidate travel schemes; adjusting model parameters of the network model based on the loss value to obtain a recommendation degree obtaining model.

7. The method of claim 6, wherein, The method further comprises: obtaining sample additional features of the sample users, wherein the sample additional features include at least one of the following features: ​ ​ a scene feature of a scenario in which the sample user generates a travel request from a sample origin to a sample destination; a user feature of the sample user; a preference feature reflecting a travel preference of the sample user; a proportion feature of different types of travel schemes that have been recommended to the sample user; a position feature representing a position of a selected travel scheme of the sample user in recommended travel schemes; the inputting of the obtained sample travel behavior features and sample scheme features into a preset network model to obtain a recommendation degree of each sample candidate travel scheme output by the network model, including: the inputting of the obtained sample travel behavior features, sample scheme features and additional features into a preset network model to obtain a recommendation degree of each sample candidate travel scheme output by the network model.

9. A travel scheme recommendation apparatus, comprising: a travel behavior feature obtaining module configured to obtain travel behavior features determined according to historical travel schemes of a target user, the travel behavior features being obtained according to interaction behaviors of the target user generated on the historical travel schemes, the interaction behaviors including a time length of browsing of the target user on each historical travel scheme and a travel mode adopted in a travel scheme finally selected by the target user; a scheme feature obtaining module configured to obtain scheme features of candidate travel schemes from an origin to a destination of the target user, the scheme features of the candidate travel schemes representing attributes of the candidate travel schemes, the attributes of the candidate travel schemes including at least one of the following attributes: price, travel time, travel distance, travel mode adopted in the candidate travel scheme, segmented walking distance and transfer times; a sub-feature obtaining unit configured to aggregate the obtained travel behavior features according to a preset aggregation dimension to obtain travel behavior sub-features; a historical travel behavior feature obtaining unit configured to, for each candidate travel scheme, calculate a matching degree of each travel behavior sub-feature with the candidate travel scheme, and obtain a historical travel behavior feature corresponding to the candidate travel scheme based on the calculated matching degrees and the travel behavior sub-features; a recommendation degree determining unit configured to determine a recommendation degree of each candidate travel scheme according to the historical travel behavior feature corresponding to the candidate travel scheme and the scheme feature of the candidate travel scheme; a travel scheme recommendation module configured to recommend each candidate travel scheme to the target user according to the recommendation degree of the candidate travel scheme.

10. The apparatus of claim 9, wherein the historical travel behavior feature obtaining unit is specifically configured to, for each candidate travel scheme, calculate a matching degree of each travel behavior sub-feature with the candidate travel scheme, and perform weighted calculation on the travel behavior sub-features with the matching degrees of the travel behavior sub-features as weights, and take a result of the weighted calculation as the historical travel behavior feature corresponding to the candidate travel scheme.

11. The apparatus of claim 9, wherein the aggregation dimension includes a time dimension of a travel scheme and / or a region dimension of a travel scheme.

12. The apparatus of any one of claims 9-11, further comprising: an accessory feature obtaining module configured to obtain accessory features of the target user, wherein the accessory features comprise at least one of the following features: a scene feature of a scene in which the target user is located when the target user generates the travel request from the starting location to the destination; a user feature of the target user; a preference feature reflecting a travel preference of the target user; a proportion feature of different types of travel schemes that have been recommended to the target user; a position feature representing a position of a travel scheme selected by the target user in recommended travel schemes; the recommendation degree determining module is specifically configured to determine the recommendation degrees of the candidate travel schemes according to the obtained travel behavior features, the scheme features and the accessory features.

13. The apparatus according to any one of claims 9-11, wherein, the recommendation degree determining module is specifically configured to input the obtained travel behavior features and the scheme features into a pre-trained recommendation degree obtaining model to obtain the recommendation degrees of the candidate travel schemes output by the recommendation degree obtaining model, wherein the recommendation degree obtaining model is a model for obtaining travel scheme recommendation degrees obtained by training a preset network model using travel behavior features and sample scheme features of sample users, and the sample scheme features are scheme features of sample candidate travel schemes from sample starting locations to sample destinations of the sample users.

14. A model training apparatus, comprising: a sample travel behavior feature obtaining module configured to obtain sample travel behavior features determined according to sample historical travel schemes of sample users, wherein the sample travel behavior features are obtained according to interaction behaviors of the sample users with respect to the sample historical travel schemes, and the interaction behaviors are durations of browsing of the sample users with respect to the sample historical travel schemes and travel modes adopted in travel schemes finally selected by the sample users; a feature and travel scheme determining module configured to determine sample scheme features of sample candidate travel schemes from sample starting locations to sample destinations of the sample users and target travel schemes selected by the sample users in the sample candidate travel schemes, wherein the scheme features of the candidate travel schemes represent attributes of the candidate travel schemes, and the attributes of the candidate travel schemes comprise at least one of the following attributes: price, travel time, travel distance, travel mode adopted in the candidate travel schemes, segmented walking distance and number of transfers; a recommendation degree obtaining module configured to input the obtained sample travel behavior features and sample scheme features into a preset network model, and the network model performs data processing in the following manner and outputs recommendation degrees of the sample candidate travel schemes: aggregating the obtained sample travel behavior features according to a preset aggregation dimension to obtain sample travel behavior sub-features; for each sample candidate travel scheme, calculating matching degrees of the sample travel behavior sub-features and the sample candidate travel scheme, and obtaining sample historical travel behavior features corresponding to the sample candidate travel scheme based on the calculated matching degrees and the sample travel behavior sub-features; and According to the sample historical travel behavior characteristics corresponding to each sample candidate travel scheme and the sample scheme characteristics of each sample candidate travel scheme, a recommendation degree of each sample candidate travel scheme is determined; a loss value generation module configured to generate a loss value for training the network model according to a numerical relationship between the recommendation degree of the target travel scheme and the recommendation degrees of other travel schemes, wherein the other travel schemes are schemes other than the target travel scheme in the sample candidate travel schemes; a model obtaining module configured to obtain a recommendation degree obtaining model by adjusting model parameters of the network model based on the loss value. 15.An electronic device, comprising: at least one processor; and a memory connected to the at least one processor in communication; wherein, the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-5 or 6-8.

16. A non-transitory computer readable storage medium having stored thereon computer instructions, wherein, The computer instructions are used to enable the computer to perform the method of any one of claims 1-5 or 6-8. 17.A computer program product comprising a computer program which, when executed by a processor, implements the method of any one of claims 1-5 or 6-8.

Images