A method and system for rearranging travel transfer plans based on a two-stage method

Abstract

A method and system for rearranging travel transfer plans based on a two-stage method, relating to the field of smart travel. In this method, a plurality of first travel transfer plans after fine-tuning and the corresponding fine-tuning scores are obtained; the plurality of the first travel transfer plans and the corresponding fine-tuning scores are input into a preset sequence generation model to obtain a plurality of first ordered sequences, the preset sequence generation model includes a plurality of generative algorithms, each generative algorithm generates one or more first ordered sequences; the plurality of the first ordered sequences are input into a context-aware preset sequence evaluation model to determine a target ordered sequence with the highest exposure purchase rate score; the target ordered sequence is displayed. Implementation of the technical solution provided in this application improves the efficiency and satisfaction of users in selecting travel plans.

Description

Technical Field

[0001] The present application relates to the technical field of smart travel, and specifically to a method and system for rearranging travel transfer plans based on a two-stage method. Background Art

[0002] With the development of intelligent transportation systems, travel plan recommendations have become an important part of improving users' travel experience.

[0003] At present, the most common way to recommend travel plans is to use precision ranking and scoring. Precision ranking and scoring are estimates of the exposure and purchase rate of each plan. If the precision ranking and scoring are used directly to sort, that is, the plan with the highest exposure and purchase rate is placed first, and then the plans are sorted down in order, the problem is that it does not consider the impact of location on the plan purchase rate, and there will be a large number of homogeneous plans (the features of homogeneous plans are basically the same, and the scores obtained are the same) piled up together, which cannot be solved.

[0004] In order to solve the above problems, this field urgently needs a more intelligent and efficient travel transfer plan rearrangement method that can comprehensively consider multiple factors and provide users with more accurate and personalized travel recommendations. Summary of the invention

[0005] The present application provides a method and system for rearranging travel transfer plans based on a two-stage method. The present application uses different types of sequence generation methods to make sequence generation more refined and intelligent, taking into account the impact of location on plan purchase rate and the mutual influence between plans. Compared with the one-stage generative model that directly generates rearranged sequences, the present application adopts a two-stage method for the first time, using an evaluation model with added context-awareness functions, which can effectively and accurately evaluate the impact of display sequences of different orders on users, and by selecting the optimal sequence, can quickly calculate and return an ordered sequence with a high score.

[0006] In a first aspect of the present application, a method for rearranging travel transfer plans based on a two-stage method is provided, which is applied to a travel plan display platform, and the method includes: Obtain multiple first trip transfer plans after refined sorting and corresponding refined sorting scores; Inputting the plurality of the first travel transfer schemes and the corresponding refined ranking scores into a preset sequence generation model to obtain a plurality of first ordered sequences, wherein the preset sequence generation model includes a plurality of generative algorithms, and each generative algorithm generates one or more first ordered sequences; Inputting a plurality of the first ordered sequences into a context-aware preset sequence evaluation model to determine a target ordered sequence with the highest exposure-to-purchase rate score; The target ordered sequence is displayed.

[0007] By adopting the above technical solutions, it can be ensured that the travel transfer solutions finally displayed to users are optimized and screened. These solutions not only take into account the original preferences and needs of users, but also further improve their rationality and convenience through algorithms. Therefore, users can find the travel solution that best meets their needs more quickly, thereby improving overall user satisfaction and experience. Using the preset sequence generation model and the context-aware sequence evaluation model, this application can accurately predict and evaluate the exposure purchase rate of different travel transfer solutions. By selecting the solution with the highest exposure purchase rate score for display, it can more effectively attract the attention of users and prompt them to make purchase decisions, thereby improving the conversion rate of the platform. Compared with the traditional travel solution display method, this application realizes the intelligent sorting and rearrangement of travel transfer solutions by introducing advanced algorithms and models. This can not only improve the intelligence level of the platform, but also enhance the competitiveness of the platform in the market by providing more personalized and accurate travel solution recommendations. In the travel solution display platform, the rational allocation of resources is the key to improving operational efficiency. Through a scientific sorting and rearrangement mechanism, this application can ensure that limited display resources are used to display the most valuable travel transfer solutions, thereby avoiding waste of resources and achieving optimal allocation of resources.

[0008] Optionally, before inputting the plurality of first ordered sequences into a context-aware preset sequence evaluation model to determine a target ordered sequence with a highest exposure-to-purchase rate score, the method further includes: Acquire characteristic data of a plurality of second travel transfer plans, the characteristic data including total duration, departure time, destination, and total price; sorting the plurality of the second travel transfer solutions according to an exposure order to obtain a second ordered sequence, wherein the exposure order refers to a time order in which the solutions are viewed by a user; assigning label data to the second travel transfer solution according to the user behavior data, and aligning the label data with the feature data of the second ordered sequence; A preset sequence evaluation model is trained according to the feature data and the label data.

[0009] By adopting the above technical solutions and introducing the feature data and user behavior data of the second travel transfer plan, the model can have a more comprehensive understanding of the attributes and user preferences of each plan, so as to make more accurate predictions. This helps to improve the accuracy and relevance of the final recommendation results. Using diversified feature data and labeled data for training can make the model better adapt to different scenarios and user groups and enhance its generalization ability. This means that the model can still maintain a high prediction accuracy when facing new travel transfer plans or users. By simulating the exposure sequence and training the model to predict the exposure purchase rate, a scientific exposure strategy basis can be provided for the platform. The platform can adjust the exposure sequence and display method according to the prediction results of the model to attract more users to click and buy. The travel transfer plans finally displayed to users are carefully screened and sorted by the model. These plans not only meet the personal preferences and needs of users, but also have a high purchase conversion rate. This helps to improve the overall experience and satisfaction of users.

[0010] Optionally, the training of a preset sequence evaluation model according to the feature data and the label data includes: Splitting a data set into training data and test data according to a preset ratio, wherein the data set includes the feature data and the label data; The exposure purchase rate of the second travel transfer plan is scored according to the training data after data processing using a preset loss function and a preset context-aware algorithm to obtain a trained preset sequence evaluation model.

[0011] By adopting the above technical solution, the data set is divided into training data and test data according to a preset ratio, which ensures that the model can maintain good performance on unseen data. This approach helps to improve the generalization ability of the model and make it more robust in practical applications. Using a preset loss function to train the model allows the model to continuously optimize its own parameters during the training process to minimize the difference between the predicted value and the actual value. This optimization process helps to improve the model's prediction accuracy of the exposure purchase rate, thereby providing users with more accurate recommendation results. The preset context-aware algorithm enables the model to consider more contextual information during the prediction process, such as the user's historical behavior, current time, geographic location, etc. The application of this algorithm can enhance the model's in-depth understanding of user needs, thereby providing travel transfer plan recommendations that are more in line with user expectations.

[0012] Optionally, the using a preset loss function and a preset context-aware algorithm to score the exposure purchase rate of the second travel transfer solution according to the training data after data processing includes: Analyze the user's travel habits based on the user's historical travel records, wherein the travel habits include the type of transfer stations and the maximum transfer time; Calculating a context score of the second travel transfer solution based on the user preference and the real-time information; The exposure-to-purchase rate score of the second travel transfer solution is determined using the processed training data, a preset loss function, and the context score.

[0013] By adopting the above technical solution and analyzing the user's historical travel records, we can deeply understand the user's travel habits, including the preferred transfer station type and the maximum acceptable transfer time. This personalized analysis provides an important basis for subsequent recommendations, making the recommendation results more in line with the actual needs of users. The context score of the second travel transfer plan is calculated based on user preferences and real-time information (such as traffic conditions, weather changes, etc.), ensuring that the recommendation results not only consider the user's static preferences, but also incorporate dynamically changing external factors. This context-awareness capability makes the recommendation system more intelligent and flexible. The exposure purchase rate score of the second travel transfer plan is determined using the processed training data, the preset loss function and the context score. This process ensures that the high-scoring plan can obtain higher exposure and purchase conversion rate when displayed through the optimization of the mathematical model, thereby improving the overall effect of the recommendation system. Through accurate recommendations and sorting, users can find travel transfer plans that meet their needs and preferences more quickly, reducing the time and energy cost of screening among many options. This efficient recommendation method helps to improve user satisfaction and loyalty. Allocating limited display resources to travel transfer plans with high exposure purchase rate scores helps to improve the overall conversion rate and revenue of the platform. At the same time, high-quality recommendation services can also attract more users to use the platform, further enhancing the platform's market competitiveness and brand influence.

[0014] Optionally, the allocating label data to the second travel transfer solution according to the user behavior data, and aligning the label data with the feature data of the second ordered sequence includes: Extracting the user's purchase record from the data source, wherein the purchase record includes the purchased plan ID and the purchase time; Matching the target second travel transfer plan according to the plan ID, and modifying the marking data of the target second travel transfer plan to a preset value; The marked data are allocated to corresponding positions of the second ordered sequence according to the scheme ID for alignment, and the aligned feature data and marked data are integrated.

[0015] By adopting the above technical solution, through the precise matching of the solution ID, it can be ensured that the user behavior data (such as purchase records) accurately corresponds to the target second travel transfer solution. This precise matching reduces the possibility of data mismatch and improves the accuracy and reliability of subsequent analysis. The marked data is assigned to the corresponding position of the second ordered sequence according to the solution ID, so as to realize the rapid alignment of the feature data and the marked data. This alignment operation not only simplifies the data processing process, but also improves the efficiency of data processing, and provides strong data support for the training and evaluation of subsequent models. The aligned and integrated feature data and marked data provide rich and accurate information for the training of the preset sequence evaluation model. This information can help the model better understand the relationship between user behavior patterns and travel transfer solutions, thereby optimizing the model's prediction ability and improving the accuracy of the exposure purchase rate score. The marked data generated based on user behavior data can reflect the user's preference and willingness to purchase for different travel transfer solutions. By incorporating this information into model training, the recommendation system can more accurately grasp user needs and provide users with more personalized travel transfer solution recommendations.

[0016] Optionally, the step of inputting the plurality of the first travel transfer solutions and the corresponding refined ranking scores into a preset sequence generation model to obtain a plurality of first ordered sequences includes: Comparing the feature data of two randomly selected second travel transfer plans to determine the similarity of the two second travel transfer plans, and when the similarity is greater than a threshold, deleting any one of the two second travel transfer plans; The above steps are repeated until the similarity of the feature data of any two second travel transfer plans is less than or equal to the threshold, and the first ordered sequence is constructed and output according to the remaining second travel transfer plans.

[0017] By adopting the above technical solution, by comparing the feature data of two randomly selected second travel transfer plans (these feature data may include time, cost, comfort, number of transfers and other factors), their similarity is calculated, and when the similarity is greater than a preset threshold, one of the plans is deleted. This process effectively reduces the redundant plans in the output sequence. Doing so not only makes the output results more refined, but also improves the user experience, because users do not need to choose from multiple very similar options. In the generated first ordered sequence, each plan is obtained according to certain standards (such as refined ranking scores) and after deduplication optimization, so the first ordered sequence can more directly reflect the priority between different plans. For users, this means that they can find the travel plan that best suits their needs faster, thereby improving the efficiency of decision-making. By presetting the sequence generation model and combining the deduplication optimization step, the system can provide users with travel suggestions in a more intelligent and personalized way. This recommendation method based on data analysis and algorithm optimization can better meet the diverse needs of different users than the traditional recommendation method based on manual experience, thereby enhancing the user experience and satisfaction. In practical applications, the deduplication method can effectively reduce the situation of recommending highly similar travel transfer plans to users. This not only improves the user experience and avoids user boredom and confusion, but also increases user trust and satisfaction with the recommendation system.

[0018] Optionally, the step of inputting the plurality of the first travel transfer solutions and the corresponding refined ranking scores into a preset sequence generation model to obtain a plurality of first ordered sequences includes: Arranging a plurality of second travel transfer schemes and defining a window of a preset size, wherein the window includes a preset number of second travel transfer schemes; The window is placed at the starting position, and the window is slid to the right according to a preset step length until the window covers all the second trip transfer solutions; At each window position, a preset number of second travel transfer plans in the window are evaluated, wherein the evaluation includes calculating the total cost, the total duration and the comfort value; An optimal solution is determined according to the evaluation result, and an output ordered sequence is constructed according to a preset number of second travel transfer solutions corresponding to the optimal solution.

[0019] By adopting the above technical solution, by defining a window of preset size and evaluating the internal travel transfer scheme at each window position, the present application can realize the refined analysis and optimization of the local scheme. This local evaluation helps to ensure that the relatively optimal travel transfer scheme combination can be found in each window. As the window slides, the entire travel transfer scheme set is gradually covered, and an ordered sequence based on the current local optimal solution is constructed at each window position. This process combines local optimization with global considerations, which helps to find a more reasonable sequence of travel transfer schemes in the global scope. Compared with a one-time evaluation of the entire travel transfer scheme set, the sliding window method significantly reduces the amount of calculation by limiting the scope of each evaluation (i.e., the number of travel transfer schemes in the window). This divide-and-conquer strategy makes the calculation process more efficient and reduces the consumption of computing resources. Since the evaluation of each window is independent, this method has the potential for parallel processing. Through parallel computing technology, the time of model training can be further shortened and the overall computing efficiency can be improved. Window size and step size are key parameters in the sliding window method, and they can be adjusted according to specific needs. By adjusting these parameters, it is possible to flexibly handle travel transfer schemes of different lengths and densities, improving the adaptability and scalability of the method. The preset sequence generation model trained by the sliding window method can more accurately predict and recommend ordered sequences of travel transfer schemes. These ordered sequences not only take into account factors such as travel cost and time, but also take into account user experience requirements such as comfort, which helps to improve user travel efficiency and satisfaction.

[0020] In a second aspect of the present application, a system for rearranging travel transfer plans based on a two-stage method is provided, characterized in that it includes a refinement module, a generation module, an evaluation module and a display module, wherein: A refined ranking module, configured to obtain a plurality of first travel transfer schemes after refined ranking and corresponding refined ranking scores; A generation module configured to input the plurality of the first travel transfer schemes and the corresponding refined ranking scores into a preset sequence generation model to obtain a plurality of first ordered sequences, wherein the preset sequence generation model includes a plurality of generation algorithms, and each generation algorithm generates one or more first ordered sequences; An evaluation module configured to input a plurality of the first ordered sequences into a context-aware preset sequence evaluation model to determine a target ordered sequence with a highest exposure-to-purchase rate score; A display module configured to display the target ordered sequence In the third aspect of the present application, an electronic device is provided, including a processor, a memory, a user interface and a network interface, the memory is used to store instructions, the user interface and the network interface are both used to communicate with other devices, and the processor is used to execute the instructions stored in the memory so that the electronic device executes any one of the methods described above.

[0021] In a fourth aspect of the present application, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores instructions, and when the instructions are executed, any of the methods described above is executed.

[0022] In summary, one or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: 1. By obtaining multiple first-trip transfer plans after fine-tuning and the corresponding fine-tuning scores, it is possible to ensure that these plans have high recommendation value at the initial stage. Subsequently, these plans and scores are input into the preset sequence generation model, and different ordered sequences are generated using a variety of generative algorithms. This method of combining fine-tuning scores and generative algorithms helps to consider users' travel needs from multiple dimensions and improve the personalization and accuracy of recommendations; 2. Input multiple first ordered sequences into a context-aware preset sequence evaluation model, which can comprehensively consider multiple contextual factors such as user historical behavior, real-time traffic conditions, weather conditions, etc., and conduct in-depth analysis and evaluation of the sequences. In this way, it can ensure that the final target ordered sequence achieves the best exposure purchase rate, that is, the travel transfer plan in the sequence is most likely to be accepted and purchased by users. This optimized display sequence can significantly improve the user experience and reduce the user's selection cost; 3. Through accurate and personalized recommendations, users can find travel transfer solutions that meet their needs more quickly, thereby improving user satisfaction. In the long run, this high-quality recommendation service can enhance users' trust and reliance on the platform and improve user loyalty; allocating limited display resources to travel transfer solutions with high exposure purchase rate scores will help improve the platform's overall conversion rate and revenue. At the same time, this resource allocation method will also help attract more users to use the platform, further enhancing the platform's market competitiveness and brand influence; 4. The application of the two-stage method encourages researchers and developers to continuously optimize and improve the sequence generation model and sequence evaluation model. By continuously trying new algorithms and strategies, the model's prediction ability and recommendation effect can be further improved, promoting the innovation and development of technology in the field of travel plan recommendation. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1It is a flowchart of a method for rearranging travel transfer plans based on a two-stage method disclosed in an embodiment of the present application; Figure 2 It is a schematic diagram of the architecture of the method for rearranging travel transfer plans based on the two-stage method disclosed in the embodiment of the present application; Figure 3 It is a module schematic diagram of a system for rearranging travel transfer plans based on a two-stage method disclosed in an embodiment of the present application; Figure 4 It is a structural schematic diagram of an electronic device disclosed in an embodiment of the present application.

[0024] Explanation of the reference numerals: 301, sorting module; 302, generation module; 303, evaluation module; 304, display module; 401, processor; 402, communication bus; 403, user interface; 404, network interface; 405, memory. DETAILED DESCRIPTION

[0025] In order to enable technicians in this field to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this application, not all of the embodiments.

[0026] In the description of the embodiments of the present application, words such as "for example" or "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "for example" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "for example" or "for example" is intended to present related concepts in a specific way.

[0027] In the description of the embodiments of the present application, the meaning of the term "multiple" refers to two or more. For example, multiple systems refer to two or more systems, and multiple screen terminals refer to two or more screen terminals. In addition, the terms "first" and "second" are used for descriptive purposes only and cannot be understood as indicating or implying relative importance or implicitly indicating the indicated technical features. Thus, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. The terms "include", "comprise", "have" and their variations all mean "including but not limited to", unless otherwise specifically emphasized.

[0028] This embodiment discloses a method for rearranging travel transfer plans based on a two-stage method, which is applied to a travel plan display platform. Figure 1 is a flow chart of a method for rearranging travel transfer plans based on a two-stage method disclosed in an embodiment of the present application, such as Figure 1 As shown, the method comprises the following steps: S110, obtaining a plurality of first trip transfer plans after refined ranking and corresponding refined ranking scores; In the initial stage of the recommendation system or search system, the main function of the recall algorithm is to quickly retrieve travel transfer solutions that may be related to the user's request from a large candidate set. These travel transfer solutions are initially screened out to form a relatively small candidate set for further processing in subsequent steps. The recall algorithm usually pursues the recall rate, that is, to find all relevant items as much as possible, although this may introduce some less relevant items.

[0029] After the recall phase, the candidate set obtained is much smaller than the original data set, but still contains more items. At this point, the task of the refined ranking algorithm is to sort these candidate items in a refined manner, evaluate each item based on multiple dimensions such as quality, relevance, and user preferences, and assign a refined ranking score to each item. The refined ranking score in the embodiment of the present application can directly reflect the exposure purchase rate.

[0030] After being processed by the refined ranking algorithm, a list of candidate solutions is obtained, which is sorted by refined ranking scores. Each solution in this list is attached with a refined ranking score. The higher the score, the higher the exposure purchase rate of the solution. These solutions may be a combination of transfers of various means of transportation such as airplanes, trains, and cars, aiming to provide users with a variety of travel options.

[0031] The embodiment of the present application uses a two-stage method to rearrange the travel transfer plans after obtaining multiple first travel transfer plans after refinement and the corresponding refined scores. The two-stage method consists of two parts: a generation sequence and an evaluation sequence. The generation sequence can use different generation algorithms, and the evaluation sequence can also use different evaluation algorithms.

[0032] S120, inputting the plurality of the first travel transfer schemes and the corresponding refined ranking scores into a preset sequence generation model to obtain a plurality of first ordered sequences, wherein the preset sequence generation model includes a plurality of generative algorithms, and each generative algorithm generates one or more first ordered sequences; The preset sequence generation model is a pre-trained model that receives the first travel transfer plan and the corresponding refined ranking score and generates an ordered sequence of travel transfer plans. The model contains a variety of generative algorithms, each of which can generate one or more ordered sequences based on the input data. Figure 2 is a schematic diagram of the architecture of the method for rearranging travel transfer plans based on a two-stage method disclosed in an embodiment of the present application, such as Figure 2As shown in the figure, sequences can be directly generated by generative algorithms or generative models. Generative algorithms include precise ranking, deduplication, beamsearch, and sliding window methods, and generative models include PRM and SGRA. Each generative algorithm generates one or more ordered sequences based on input data.

[0033] Optionally, the step of inputting the plurality of the first travel transfer solutions and the corresponding refined ranking scores into a preset sequence generation model to obtain a plurality of first ordered sequences includes: Comparing the feature data of two randomly selected second travel transfer plans to determine the similarity of the two second travel transfer plans, and when the similarity is greater than a threshold, deleting any one of the two second travel transfer plans; The above steps are repeated until the similarity of the feature data of any two second travel transfer plans is less than or equal to the threshold, and an output ordered sequence is constructed according to the remaining second travel transfer plans.

[0034] Define the calculation rules of similarity. In the embodiment of the present application, the similarity of two travel transfer plans is determined by features, including departure station, fare and transit time. The similarity can be calculated by the proportion of the same items in these features. For example: if the two plans are the same in all three features (departure station, fare, transit time), the similarity is 1 (or 100%); if the two plans are the same in two features (such as the departure station and fare are the same, but the transit time is different), the similarity is 2 / 3 (or about 67%); if the two plans are different in all three features, the similarity is 0.

[0035] Randomly extract two second travel transfer solutions from the data set. These solutions contain the feature data defined above. Compare the feature data of the two extracted travel transfer solutions, and calculate the similarity of the two solutions according to the similarity calculation rules defined previously. If the calculated similarity is greater than the preset threshold (such as 0.6 or 60%), it is considered that the two solutions are too similar in content and one of them needs to be removed to avoid redundancy. The travel transfer solution to be deleted can be determined according to the actual situation, such as retaining the solution with a higher refined ranking score or randomly deleting it. If the similarity is less than or equal to the threshold, the two solutions are retained unchanged. Continue to randomly extract new solution pairs from the data set, repeat the above steps of calculating similarity and applying deduplication logic until the similarity of any two travel transfer solutions in the data set is less than or equal to the threshold. This process may require multiple iterations to ensure that all redundant solutions are effectively removed. After the deduplication process is completed, the final output ordered sequence is constructed based on the remaining travel transfer solutions. This sequence can be sorted according to the refined ranking score or other user preference factors for recommendation to users.

[0036] By comparing and removing travel transfer options with similarity above a threshold, deduplication ensures that the final recommendation list does not contain a large number of options with similar content. This helps users see more different types of travel options and improves the diversity of recommendations. Users may feel confused and dissatisfied when faced with a large number of similar options because similar options do not provide them with more choice value. Deduplication improves user experience by reducing redundant options, allowing users to find travel options that meet their needs more quickly. In a recommendation system, processing a large number of similar options consumes additional computing resources. Deduplication reduces the computational burden of the system and improves recommendation efficiency by reducing the number of options that need to be evaluated. When the similarity threshold is set appropriately, deduplication can retain the most representative and competitive travel transfer options. These options perform well in multiple dimensions (such as cost, duration, and comfort), so they are more likely to meet user needs, thereby enhancing the accuracy of recommendations.

[0037] Optionally, the step of inputting the plurality of the first travel transfer solutions and the corresponding refined ranking scores into a preset sequence generation model to obtain a plurality of first ordered sequences includes: Arranging a plurality of second travel transfer schemes and defining a window of a preset size, wherein the window includes a preset number of second travel transfer schemes; The window is placed at the starting position, and the window is slid to the right according to a preset step length until the window covers all the second trip transfer solutions; At each window position, a preset number of second travel transfer plans in the window are evaluated, wherein the evaluation includes calculating the total cost, the total duration and the comfort value; An optimal solution is determined according to the evaluation result, and an output ordered sequence is constructed according to a preset number of second travel transfer solutions corresponding to the optimal solution.

[0038] For example, defining a window size of 3 means that 3 consecutive transfer plans will be considered simultaneously. Assume that the transfer plans in the ordered sequence are as follows (for simplicity, only some key information is listed here, such as departure time, transfer city, arrival time and price): Plan 1: A->C->B, departure time 08:00, arrival time 15:00, price 300; Plan 2: A->D->B, departure time 09:00, arrival time 14:30, price 400; Plan 3: A->E->B, departure time 09:30, arrival time 16:00, price 350; Plan 4: A->F->B, departure time 10:00, arrival time 15:45, price 380; ... (more solutions) The window starts at the beginning of the sequence, covering options 1, 2, and 3. Then, the window slides to the right one position at a time, covering options 2, 3, and 4, followed by options 3, 4, and the next option, until the window covers the entire sequence. At each window position, the system evaluates the sequence of options within the window. The evaluation may include calculating the total travel time, total fare, number of transfers, flight connection time, etc. During the sliding process, the system records the window position with the best evaluation result and its corresponding sequence. In this example, the optimal solution may be the sequence with the shortest total travel time, the fewest transfers, or the highest comprehensive score. Suppose that during the sliding process, the system finds that when the window covers options X, Y, and Z, the sequence composed of these three options achieves the best balance in terms of total travel time, fare, and transfer comfort. After the sliding window method is completed, the system outputs the recorded optimal solution as the final ordered sequence, that is, the sequence of transfer options that the user should choose.

[0039] By defining a window of preset size and evaluating the travel transfer schemes within the window at each window position, the local optimal solution can be quickly obtained. This local evaluation method avoids the high computational cost of global search, making the algorithm more efficient in practical applications. As the window slides, the algorithm gradually traverses all solutions and finds the local optimal solution at each window position. The combination of these local optimal solutions ultimately helps to build a global or near-global optimal travel solution sequence. The size of the preset window can be adjusted according to specific needs to adapt to data sets of different sizes and different optimization goals. For example, when the data volume is large and the computing resources are limited, a smaller window can be selected to reduce the amount of calculation; while in the case of pursuing higher accuracy, a larger window can be selected to include more scheme combinations. The step size setting also provides flexibility, allowing the algorithm to move the window at different speeds during the traversal process. A smaller step size can capture more detailed information, but will increase the amount of calculation; a larger step size can increase the calculation speed, but may miss some important local optimal solutions. At each window position, the algorithm evaluates the travel transfer schemes within the window in multiple dimensions, including total cost, total duration, and comfort value. This comprehensive evaluation method helps to consider users' needs and preferences more comprehensively, thereby generating a travel plan sequence that better meets users' expectations.

[0040] Alternatively, the first ordered sequence can be generated based on beamsearch. Specifically, a suitable beam width B is selected, which limits the number of candidate solutions in each layer (or step). Based on the initial conditions (such as the departure and destination), an initial set of candidate travel transfer solutions is generated, and the initial evaluation value (such as cost, time, etc.) of each solution is calculated. Select the B best solutions from the initial candidate set as the initial beam. For each solution in the beam, generate its possible next step (or next station) expansion solution. This may involve choosing different means of transportation, transfer stations, etc. Evaluate each newly generated solution and calculate its total evaluation value (for example, total travel time plus possible additional costs). Before adding a new solution to the beam, check whether it is too similar to the solutions already in the beam (using the similarity calculation method you mentioned). If the similarity is greater than the threshold, discard the solution or replace the old solution with the highest similarity in the beam (depending on the specific implementation). From all newly generated solutions, select the B best solutions (based on the evaluation value) as the new beam. If any solution in the beam reaches the end point (such as the destination), the algorithm can terminate early. Set a maximum number of iterations, and stop expanding when this number is reached. If the solutions in the beam do not change significantly after several consecutive iterations (such as the change in the evaluation value is less than a certain threshold), stop iterating. Select the solution with the best evaluation value from the final beam as the final output. If necessary, you can sort the solutions in the beam according to the evaluation value or other criteria to build an ordered sequence.

[0041] Beam Search effectively reduces the size of the search space by limiting the number of nodes expanded at each step (i.e., beam width), thereby reducing the amount of computation. Compared with exhaustive search, beam search can find a near-optimal solution in a reasonable time. Since beam search retains the current optimal candidate solution at each step, the quality of the candidate solution gradually improves as the search proceeds, and the algorithm can converge to the optimal solution or a near-optimal solution faster. Unlike the greedy algorithm, beam search retains multiple candidate solutions at each step instead of just one optimal solution. The advantage of this is that more paths can be explored during the search process, thereby increasing the possibility of finding the global optimal solution. Beam width is a key parameter in the beam search algorithm, which can be adjusted according to the complexity of the actual problem and the limitations of computing resources. A smaller beam width can speed up the search but may sacrifice the quality of the solution; a larger beam width can improve the quality of the solution but increase the amount of computation. Therefore, by adjusting the beam width, the algorithm can strike a balance between search efficiency and solution quality. The beam search algorithm is not only suitable for the optimization problem of travel transfer solutions, but can also be widely used in other scenarios that require sequence generation and optimization, such as machine translation, speech recognition, text summarization, etc. Its flexibility and adaptability make Beam Search a very useful heuristic search algorithm. Due to its high search efficiency, Beam Search can provide users with high-quality travel transfer solutions in a short time. This is very important for users who need to make decisions quickly. By retaining multiple candidate solutions and removing duplicates, Beam Search can provide users with a variety of travel options. Users can choose the most suitable solution according to their needs and preferences.

[0042] S130, inputting the plurality of first ordered sequences into a context-aware preset sequence evaluation model to determine a target ordered sequence with the highest exposure-to-purchase rate score; The first ordered sequence is generated by different algorithms or strategies, and each sequence contains a series of elements arranged in a specific order. These sequences are input into a context-aware preset sequence evaluation model. Figure 2As shown in the figure, the context-aware sequence evaluation model can be a single / bidirectional LSTM or MSA with / without position embedding (multi-head self-attention mechanism with position / without position embedding), etc. This model has the ability to process context information, that is, it can understand and consider the impact of the user's current situation, preferences, historical behavior, and external environmental factors (such as time, location, weather, etc.) on sequence evaluation. By introducing context information, the model can more accurately evaluate the potential value and effect of the sequence. In the model, each input first ordered sequence is evaluated for its exposure to purchase rate score. The exposure to purchase rate is an important indicator to measure the effect of a sequence. It reflects the ability of the elements in the sequence to be converted into actual user behaviors such as purchases or clicks after being exposed. The model calculates this score based on the content, order, and context information of the sequence. The model selects the one with the highest exposure to purchase rate score from all the input first ordered sequences as the target ordered sequence. This sequence is considered to be the sequence that is most likely to bring high conversion rate and user satisfaction in the current context.

[0043] Optionally, before inputting the plurality of first ordered sequences into a context-aware preset sequence evaluation model to determine a target ordered sequence with a highest exposure-to-purchase rate score, the method further includes: Acquire characteristic data of a plurality of second travel transfer plans, the characteristic data including total duration, departure time, destination, and total price; sorting the plurality of the second travel transfer solutions according to an exposure order to obtain a second ordered sequence, wherein the exposure order refers to a time order in which the solutions are viewed by a user; assigning label data to the second travel transfer solution according to the user behavior data, and aligning the label data with the feature data of the second ordered sequence; A preset sequence evaluation model is trained according to the feature data and the label data.

[0044] Acquire feature data of multiple second travel transfer plans, which fully describe the attributes of each plan, including but not limited to total duration, departure time, destination and total price (i.e., the feature data structure of each single plan is [X1, X2, X3, ..., Xn], where n is the number of features). These features together form the basis for user decision-making. Sort multiple second travel transfer plans according to the exposure order (i.e., the time order in which the user sees these plans), so as to construct a second ordered sequence. The second ordered sequence can be expressed as {[X1, X2, X3, ..., Xn] for the first-ranked plan, [X1, X2, X3, ..., Xn] for the second-ranked plan, ..., [X1, X2, X3, ..., Xn] for the m-ranked plan}, where m is the number of plans in the sequence. Assign label data to each travel transfer plan in the second ordered sequence based on user behavior data (such as click, browse, purchase and other behavior records). The structure of the labeled data is [Y for the 1st ranked solution, Y for the 2nd ranked solution, ..., Y for the mth ranked solution], where Y is the label value, for example, 1 means that the solution is purchased, and 0 means that the solution is exposed but not purchased. This step ensures that the model can learn the actual feedback of users on different solutions. Strictly aligning the assigned labeled data with the feature data of the second ordered sequence to ensure that the features of each solution correspond one-to-one with its corresponding user behavior tags is the key to the accuracy of model training. Use the above-constructed feature data and labeled data to train the preset sequence evaluation model. This training process aims to enable the model to learn the mutual influence between different solutions in the sequence and the user behavior patterns, so as to predict and optimize the exposure purchase rate of the sequence.

[0045] By obtaining the feature data of multiple second travel transfer plans, including key information such as total duration, departure time, destination and total price, the richness and accuracy of the model training data are ensured. These feature data comprehensively reflect all aspects of the travel plan and provide sufficient decision-making basis for the model. The travel transfer plans are sorted according to the exposure order to obtain the second ordered sequence. This step enables the model to learn the impact of the order relationship between different plans in the sequence on user decisions, thereby improving the model's sensitivity to the sequence order. According to user behavior data, label data is assigned to the travel transfer plan, and these label data are aligned with the feature data. This step enables the model to directly use the user's actual feedback to optimize its prediction ability. Through continuous learning and adjustment, the model can more accurately predict which plans are more likely to be purchased by users, thereby improving the accuracy of recommendations or displays. Using carefully prepared feature data and label data to train the preset sequence evaluation model can significantly improve the performance of the model. Through training, the model can learn complex feature combinations and sequence patterns, thereby more accurately evaluating the exposure purchase rate scores of different ordered sequences. This enables the model to more quickly find the target ordered sequence with the highest exposure purchase rate score in practical applications, improving overall efficiency and effectiveness. The preset sequence evaluation model is designed to be context-aware, which means that it can evaluate the quality of ordered sequences by taking into account the user's current context (such as time, location, user preferences, etc.). Through data preparation and model optimization during the training process, the model's context-awareness can be further enhanced, enabling it to more accurately reflect the user's decision-making behavior in different situations.

[0046] Optionally, the training of a preset sequence evaluation model according to the feature data and the label data includes: Splitting a data set into training data and test data according to a preset ratio, wherein the data set includes the feature data and the label data; The exposure purchase rate of the second travel transfer plan is scored according to the training data after data processing using a preset loss function and a preset context-aware algorithm to obtain a trained preset sequence evaluation model.

[0047] The dataset containing feature data and labeled data is split into training data and test data according to a preset ratio (such as 8:2). This step is to ensure that the model can learn enough data features during the training process and evaluate its generalization ability during the testing phase. A series of preprocessing operations are performed on the feature data in the training data to improve the training efficiency and performance of the model. These preprocessing operations include but are not limited to: Missing value handling: For missing values ​​in the data, adopt appropriate filling strategies, such as using the mean, median, mode, or specific default values ​​to ensure that each feature has a complete value for the model to learn.

[0048] Normalization of continuous features: In order to eliminate the impact of different dimensions on model training, the continuous features are normalized. This can be achieved through methods such as MinMaxScaler and z-score, so that all continuous features are scaled to the same scale.

[0049] Discrete feature encoding processing: For discrete features, use appropriate encoding methods (such as One-Hot encoding) to convert them into numerical representations that the model can process. In this way, the model can learn the differences and connections between discrete features.

[0050] After data processing, the preset sequence evaluation model is trained using a preset loss function (such as logloss, which is often used to evaluate classification problems, especially when dealing with unbalanced data sets) and a preset context-aware algorithm (such as Multihead Self Attention (MSA) with / without position embedding or unidirectional and bidirectional LSTM). During the training process, the processed training data is used to score the exposure-to-purchase rate of each travel transfer scheme in the sequence through a context-aware algorithm. This step is designed to enable the model to learn the mutual influence between different schemes in the sequence and the user behavior patterns, so that the exposure-to-purchase rate of each scheme can be predicted. Finally, after multiple iterations of training, a trained preset sequence evaluation model is obtained. The model can generate an exposure-to-purchase rate score for each scheme in the sequence based on the input ordered sequence feature data, so as to determine the target ordered sequence with the highest exposure-to-purchase rate score in practical applications.

[0051] By splitting the data set into training data and test data according to a preset ratio (such as 8:2), it is ensured that the model can learn enough data features during the training process, and its performance on new data can be evaluated during the test phase. This splitting method helps to improve the generalization ability of the model, that is, the model's ability to predict unseen data. Preprocessing the feature data in the training data, such as missing value processing, continuous feature normalization processing, and discrete feature encoding processing, can effectively eliminate noise and inconsistency in the data and improve data quality. These preprocessing steps help optimize the model training process and improve the performance and stability of the model. Using a preset context-aware algorithm to score the exposure purchase rate of each travel transfer plan in the sequence can fully consider the mutual influence between different plans in the sequence and user behavior patterns. This algorithm can capture the temporal information and contextual relationships in the sequence, thereby improving the model's understanding and processing capabilities of sequence data. Using a preset loss function as the objective function of model training can ensure that the model can continuously optimize its prediction ability during the training process and reduce the error between the predicted value and the actual value. The trained preset sequence evaluation model is obtained through the training process and can be verified and evaluated on the test data. This process not only helps to verify the training effect of the model, but also provides data support for subsequent model optimization. At the same time, by quantitatively evaluating the performance indicators of the model (such as accuracy, recall rate, F1 score, etc.), we can more intuitively understand the advantages and disadvantages of the model and provide a reference for subsequent application and improvement.

[0052] Optionally, the using a preset loss function and a preset context-aware algorithm to score the exposure purchase rate of the second travel transfer solution according to the training data after data processing includes: Analyze the user's travel habits based on the user's historical travel records, wherein the travel habits include the type of transfer stations and the maximum transfer time; Calculating a context score of the second travel transfer solution based on the user preference and the real-time information; The exposure-to-purchase rate score of the second travel transfer solution is determined using the processed training data, a preset loss function, and the context score.

[0053] In-depth analysis is performed based on the user's historical travel records to reveal the user's travel habits. These habits include but are not limited to the types of transfer stations preferred by the user (such as large transportation hubs, transfer stations in specific cities, etc.) and the maximum transfer time (i.e., the longest transfer time the user is willing to spend on the journey). This step is the key to understanding the user's needs and preferences, and provides an important basis for the subsequent calculation of context scores. Based on the user's preferences and real-time information (such as current weather, traffic conditions, special events, etc.), the platform will calculate the context score for each second-trip transfer plan. This step is to evaluate the applicability and attractiveness of the plan in the current situation. For example, if the weather at a transfer station is bad (such as rain), then the plan that includes the transfer station may receive a lower context score. On the contrary, if a plan can make full use of the user's preferences (such as the user's preference to transfer in a specific city) and is consistent with real-time information (such as the current smooth traffic in the city), then the plan may receive a higher context score. After obtaining the context score of each second-trip transfer plan, the platform will combine the processed training data, the preset loss function and these context scores to determine the exposure purchase rate score of each plan. This step is the core of model training, in which the preset loss function is used to guide the model optimization process to ensure that the model can accurately predict the exposure purchase rate of each plan. Specifically, the model will learn how to evaluate the attractiveness of different plans based on feature data (such as total duration, departure time, arrival place, and total price, etc.) and context scores, and give corresponding exposure purchase rate scores. The exposure purchase rate score of each ordered sequence is the average exposure purchase rate of each plan in the ordered sequence. Feature data is the basic information used by the model to evaluate the attractiveness of the plan, including but not limited to: Total duration: The total time required to complete the entire trip, which is one of the important considerations when users choose a plan. Departure time: The starting time of the trip, which may affect the user's travel arrangements and comfort. Destination: The destination of the trip, although fixed in this scenario, may be variable in a wider scenario (such as multi-destination travel planning). Total price: The total cost required to complete the trip, which is a key economic factor in the user's decision-making process.

[0054] In addition to feature data, the model may also consider contextual information to give additional scores. The context score may be based on a variety of factors, such as: User preferences: historical behavior and preference settings of users, used for personalized recommendations. Real-time traffic conditions: current road congestion, public transportation schedules, etc., which affect travel time and comfort. Weather conditions: weather conditions at the destination, which may affect users' travel willingness and comfort.

[0055] The context score is a supplement to the feature data evaluation, which enables the model to more accurately reflect the actual situation and user needs. The model uses feature data and context scores to evaluate the attractiveness of each travel transfer solution through complex algorithms (such as machine learning models, deep learning networks, etc.). This evaluation process may involve multiple steps, including feature extraction, weight assignment, nonlinear transformation, etc., and finally outputs a numerical score representing the attractiveness of the solution. The exposure purchase rate score is a predicted value given by the model based on the attractiveness of the solution, indicating the probability that the user may purchase (or choose the solution) when the solution is presented to the user. This score is a quantitative evaluation of the market potential of the solution by the model, which is of great significance for optimizing the recommendation list and improving the conversion rate. After generating an ordered sequence, the model calculates the exposure purchase rate score of the sequence. This score is the average of the exposure purchase rate scores of all solutions in the sequence, reflecting the attractiveness of the entire sequence to the user. By comparing the exposure purchase rate scores of different sequences, the model can optimize the recommendation strategy and improve user satisfaction.

[0056] By analyzing the user's historical travel records, including travel habits such as transfer station type and maximum transfer time, the platform can understand the user's preferences and needs more accurately. This personalized analysis makes the recommended transfer plan closer to the user's actual usage scenario, thereby improving user satisfaction and purchase conversion rate. The context score is calculated by combining user preferences and real-time information (such as weather, traffic conditions, flight / train delays, etc.), so that the recommended transfer plan can dynamically adapt to changes in the external environment. This real-time performance not only improves the user experience, but also reduces travel inconvenience caused by unforeseen factors. Through the preset loss function and context score, the platform can quantitatively evaluate multiple candidate transfer plans and determine their exposure purchase rate scores. This scoring mechanism provides service providers with a scientific decision-making basis, helping them optimize resource allocation, improve operational efficiency and profitability. In the highly competitive travel service market, being able to provide more personalized, real-time and accurate transfer plan recommendations will significantly enhance the market competitiveness of service providers. This will not only help attract new users, but also enhance the loyalty of old users and promote word-of-mouth communication.

[0057] Optionally, the allocating label data to the second travel transfer solution according to the user behavior data, and aligning the label data with the feature data of the second ordered sequence includes: Extracting the user's purchase record from the data source, wherein the purchase record includes the purchased plan ID and the purchase time; Matching the target second travel transfer plan according to the plan ID, and modifying the marking data of the target second travel transfer plan to a preset value; The marked data are allocated to corresponding positions of the second ordered sequence according to the scheme ID for alignment, and the aligned feature data and marked data are integrated.

[0058] Extract the user's purchase records from data sources (such as databases, log files, etc.). These purchase records contain the plan ID (each travel plan has a unique identifier) ​​and the purchase time that the user actually purchased. This information is the key basic data for understanding user behavior and evaluating the attractiveness of travel plans. According to the extracted plan ID, match the target second travel transfer plan in the existing travel transfer plan database. When the corresponding plan is found, the platform will modify the tag data of the plan to a preset value. This preset value is usually used to represent a certain behavior or attitude of the user towards the plan, such as purchase, click, browse, etc. The specific value is set according to business needs and data analysis goals. The platform assigns the modified tag data to the corresponding position of the second ordered sequence according to the plan ID. The alignment operation ensures that each tag data can be associated with the correct feature data, providing an accurate basis for subsequent data analysis and model training. Finally, the platform integrates the aligned feature data and tag data. This step is to unify data from different sources and in different formats into a complete data set for further data analysis, model training or business decision-making. The integrated dataset not only contains detailed information on travel transfer options (such as departure point, destination, transfer station, transfer time, etc.), but also includes users' actual behavioral feedback on these options (such as purchase status), which is crucial for evaluating and optimizing travel options.

[0059] By extracting the user's purchase record from the data source, including the purchased plan ID and purchase time, the user's actual purchase behavior can be accurately matched with a specific travel transfer plan. This precise matching ensures that the basic data for subsequent analysis is accurate. As a direct reflection of user behavior, the purchase record can reflect the user's preference and satisfaction with the travel transfer plan in real time. Using this data to mark the second travel transfer plan can ensure the real-time and effectiveness of the recommendation system. By analyzing the user's purchase record, the user's travel preferences, such as the preferred transfer station type, transfer time, etc., can be identified. This information is crucial for optimizing the personalized recommendation algorithm. Aligning the labeled data (such as a preset value) with the feature data of the second ordered sequence and integrating them together can further enrich the data dimension of the recommendation system. This enables the recommendation system to consider the user's personalized needs more comprehensively, thereby improving the accuracy and relevance of the recommendation results. The labeled data is directly assigned to the corresponding position of the second ordered sequence through the plan ID, which simplifies the data processing process and reduces the complexity and error rate of data processing. The integration of labeled data and feature data helps the recommendation system better understand the user's behavior pattern and optimize the algorithm model. This will improve the prediction accuracy and response speed of the system and enhance the user experience.

[0060] S140: Display the target ordered sequence.

[0061] For each travel plan in the sequence, its relevant information needs to be presented in an appropriate manner. This can include text descriptions, icons, charts, and other forms. For example, different colored backgrounds or borders can be used to distinguish the exposure purchase rate scores of different plans; icons can be used to indicate the type or location of transfer stations; charts can be used to show the comparative relationship between plans, etc. In order to improve the user experience, some interactive functions can also be added to the display interface. For example, allow users to click on a travel plan to view more detailed information; provide filtering and sorting functions to help users quickly locate the plan of interest; set reminders or collection functions for users to refer to in the future, etc. The display interface should have good adaptability and responsiveness, and be able to adjust adaptively according to the screen size and resolution of different devices (such as mobile phones, tablets, computers, etc.) to ensure that users can get a good browsing experience on different devices. Finally, in order to evaluate the display effect and continuously optimize the user experience, some feedback mechanisms can also be set up. For example, through user surveys or behavioral data analysis to understand users' satisfaction with the display interface and improvement suggestions; adjust the interface design or functional layout according to user feedback, etc.

[0062] This embodiment also discloses a system for rearranging travel transfer plans based on a two-stage method. Figure 3 is a schematic diagram of a module of a system for rearranging travel transfer plans based on a two-stage method disclosed in an embodiment of the present application, such as Figure 3 As shown, the system includes a sorting module 301, a generating module 302, an evaluating module 303 and a display module 304, wherein: A refined ranking module 301 is configured to obtain a plurality of refined first travel transfer schemes and corresponding refined ranking scores; A generating module 302 is configured to input the plurality of the first travel transfer schemes and the corresponding refined ranking scores into a preset sequence generating model to obtain a plurality of first ordered sequences, wherein the preset sequence generating model includes a plurality of generating algorithms, and each generating algorithm generates one or more first ordered sequences; An evaluation module 303 is configured to input a plurality of the first ordered sequences into a context-aware preset sequence evaluation model to determine a target ordered sequence with the highest exposure-to-purchase rate score; The display module 304 is configured to display the target ordered sequence.

[0063] Optionally, the system further comprises a training module, wherein the training module is configured to: Acquire characteristic data of a plurality of second travel transfer plans, the characteristic data including total duration, departure time, destination, and total price; sorting the plurality of the second travel transfer solutions according to an exposure order to obtain a second ordered sequence, wherein the exposure order refers to a time order in which the solutions are viewed by a user; assigning label data to the second travel transfer solution according to the user behavior data, and aligning the label data with the feature data of the second ordered sequence; A preset sequence evaluation model is trained according to the feature data and the label data.

[0064] Optionally, the training module is configured to: Splitting a data set into training data and test data according to a preset ratio, wherein the data set includes the feature data and the label data; The exposure purchase rate of the second travel transfer plan is scored according to the training data after data processing using a preset loss function and a preset context-aware algorithm to obtain a trained preset sequence evaluation model.

[0065] Optionally, the training module is configured to: Analyze the user's travel habits based on the user's historical travel records, wherein the travel habits include the type of transfer stations and the maximum transfer time; Calculating a context score of the second travel transfer solution based on the user preference and the real-time information; The exposure-to-purchase rate score of the second travel transfer solution is determined using the processed training data, a preset loss function, and the context score.

[0066] Optionally, the training module is configured to: Extracting the user's purchase record from the data source, wherein the purchase record includes the purchased plan ID and the purchase time; Matching the target second travel transfer plan according to the plan ID, and modifying the marking data of the target second travel transfer plan to a preset value; The marked data are allocated to corresponding positions of the second ordered sequence according to the scheme ID for alignment, and the aligned feature data and marked data are integrated.

[0067] Optionally, the generation module is configured to: Comparing the feature data of two randomly selected second travel transfer plans to determine the similarity of the two second travel transfer plans, and when the similarity is greater than a threshold, deleting any one of the two second travel transfer plans; The above steps are repeated until the similarity of the feature data of any two second travel transfer plans is less than or equal to the threshold, and the first ordered sequence is constructed and output according to the remaining second travel transfer plans.

[0068] Optionally, the generation module is configured to: Arranging a plurality of second travel transfer schemes and defining a window of a preset size, wherein the window includes a preset number of second travel transfer schemes; The window is placed at the starting position, and the window is slid to the right according to a preset step length until the window covers all the second trip transfer solutions; At each window position, a preset number of second travel transfer plans in the window are evaluated, wherein the evaluation includes calculating the total cost, the total duration and the comfort value; An optimal solution is determined according to the evaluation result, and a first ordered sequence is constructed and outputted according to a preset number of second travel transfer solutions corresponding to the optimal solution.

[0069] It should be noted that: when the device provided in the above embodiment realizes its function, only the division of the above functional modules is used as an example. In actual application, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the device and method embodiments provided in the above embodiment belong to the same concept, and the specific implementation process is detailed in the method embodiment, which will not be repeated here.

[0070] This embodiment also discloses an electronic device, referring to Figure 4 The electronic device may include: at least one processor 401 , at least one communication bus 402 , a user interface 403 , a network interface 404 , and at least one memory 405 .

[0071] The communication bus 402 is used to realize the connection and communication between these components.

[0072] The user interface 403 may include a display screen (Display) and a camera (Camera). Optionally, the user interface 403 may also include a standard wired interface and a wireless interface.

[0073] The network interface 404 may optionally include a standard wired interface or a wireless interface (such as a WI-FI interface).

[0074] Among them, the processor 401 may include one or more processing cores. The processor 401 uses various interfaces and lines to connect various parts in the entire server, and executes various functions of the server and processes data by running or executing instructions, programs, code sets or instruction sets stored in the memory 405, and calling data stored in the memory 405. Optionally, the processor 401 can be implemented in at least one hardware form of digital signal processing (Digital Signal Processing, DSP), field programmable gate array (Field-Programmable Gate Array, FPGA), and programmable logic array (Programmable Logic Array, PLA). The processor 401 can integrate one or a combination of a central processing unit (Central Processing Unit, CPU), a graphics processing unit (Graphics Processing Unit, GPU) and a modem. Among them, the CPU mainly processes the operating system, user interface and application programs; the GPU is responsible for rendering and drawing the content to be displayed on the display screen; the modem is used to process wireless communications. It can be understood that the above-mentioned modem may not be integrated into the processor 401, and it can be implemented separately through a chip.

[0075] Among them, the memory 405 may include a random access memory (Random Access Memory, RAM) and may also include a read-only memory (Read-Only Memory). Optionally, the memory 405 includes a non-transitory computer-readable storage medium. The memory 405 can be used to store instructions, programs, codes, code sets or instruction sets. The memory 405 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as a touch function, a sound playback function, an image playback function, etc.), instructions for implementing the above-mentioned various method embodiments, etc.; the data storage area may store data involved in the above-mentioned various method embodiments, etc. The memory 405 may also be optionally at least one storage device located away from the aforementioned processor 401. As Figure 4 As shown, the memory 405 as a computer storage medium may include an operating system, a network communication module, a user interface module, and an application program for a method for rearranging travel transfer plans based on a two-stage method.

[0076] exist Figure 4 In the electronic device shown, the user interface 403 is mainly used to provide an input interface for the user and obtain data input by the user; and the processor 401 can be used to call an application program stored in the memory 405 for a method for rearranging travel transfer plans based on a two-stage method. When executed by one or more processors 401, the electronic device executes one or more methods in the above-mentioned embodiments.

[0077] It should be noted that, for the aforementioned method embodiments, for the sake of simplicity, they are all described as a series of action combinations, but those skilled in the art should be aware that the present application is not limited by the order of the actions described, because according to the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required for the present application.

[0078] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0079] In the several embodiments provided in this application, it should be understood that the disclosed devices can be implemented in other ways. For example, the device embodiments described above are only schematic, such as the division of units, which is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some service interfaces, and the indirect coupling or communication connection of devices or units can be electrical or other forms.

[0080] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0081] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0082] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable memory. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a memory 405, including a number of instructions for a computer device (which can be a personal computer, server or network device, etc.) to perform all or part of the steps of the various embodiments of the present application. The aforementioned memory 405 includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a magnetic disk or an optical disk.

[0083] The above is only an exemplary embodiment of the present disclosure and cannot be used to limit the scope of the present disclosure. That is, any equivalent changes and modifications made according to the teachings of the present disclosure are still within the scope of the present disclosure. After considering the disclosure of the specification, those skilled in the art will easily think of other embodiments of the present disclosure. This application is intended to cover any modification, use or adaptation of the present disclosure, which follows the general principles of the present disclosure and includes common knowledge or customary technical means in the technical field that are not recorded in the present disclosure. The description and examples are only regarded as exemplary, and the scope and spirit of the present disclosure are defined by the claims.

Claims

1. A method for rearranging travel transfer plans based on a two-stage method, characterized in that: Applied to a travel plan display platform, the method includes: Obtain multiple first trip transfer plans after refined sorting and corresponding refined sorting scores; Inputting the plurality of the first travel transfer schemes and the corresponding refined ranking scores into a preset sequence generation model to obtain a plurality of first ordered sequences, wherein the preset sequence generation model includes a plurality of generative algorithms, and each generative algorithm generates one or more first ordered sequences; Inputting a plurality of the first ordered sequences into a context-aware preset sequence evaluation model to determine a target ordered sequence with the highest exposure-to-purchase rate score; The target ordered sequence is displayed.

2. The method for rearranging travel transfer plans based on the two-stage method according to claim 1, characterized in that: Before inputting the plurality of first ordered sequences into a context-aware preset sequence evaluation model to determine a target ordered sequence with a highest exposure-to-purchase rate score, the method further includes: Acquire characteristic data of a plurality of second travel transfer plans, the characteristic data including total duration, departure time, destination, and total price; sorting the plurality of the second travel transfer solutions according to an exposure order to obtain a second ordered sequence, wherein the exposure order refers to a time order in which the solutions are viewed by a user; assigning label data to the second travel transfer solution according to the user behavior data, and aligning the label data with the feature data of the second ordered sequence; A preset sequence evaluation model is trained according to the feature data and the label data.

3. The method for rearranging travel transfer plans based on the two-stage method according to claim 2, characterized in that: The training of the preset sequence evaluation model according to the feature data and the label data comprises: Splitting a data set into training data and test data according to a preset ratio, wherein the data set includes the feature data and the label data; The exposure purchase rate of the second travel transfer plan is scored according to the training data after data processing by using a preset loss function and a preset context-aware algorithm to obtain a trained preset sequence evaluation model.

4. The method for rearranging travel transfer plans based on the two-stage method according to claim 3, characterized in that: The step of using a preset loss function and a preset context-aware algorithm to score the exposure purchase rate of the second travel transfer solution according to the training data after data processing includes: Analyze the user's travel habits based on the user's historical travel records, wherein the travel habits include the type of transfer stations and the maximum transfer time; Calculating a context score of the second travel transfer solution based on the user preference and the real-time information; The exposure-to-purchase rate score of the second travel transfer solution is determined using the processed training data, a preset loss function, and the context score.

5. The method for rearranging travel transfer plans based on the two-stage method according to claim 2, characterized in that: The allocating label data to the second travel transfer scheme according to the user behavior data, and aligning the label data with the feature data of the second ordered sequence includes: Extracting the user's purchase record from the data source, wherein the purchase record includes the purchased plan ID and the purchase time; Matching the target second travel transfer plan according to the plan ID, and modifying the marking data of the target second travel transfer plan to a preset value; The marked data are allocated to corresponding positions of the second ordered sequence according to the scheme ID for alignment, and the aligned feature data and marked data are integrated.

6. The method for rearranging travel transfer plans based on the two-stage method according to claim 2, characterized in that: The step of inputting the plurality of the first travel transfer solutions and the corresponding refined ranking scores into a preset sequence generation model to obtain a plurality of first ordered sequences comprises: Comparing the feature data of two randomly selected second travel transfer plans to determine the similarity of the two second travel transfer plans, and when the similarity is greater than a threshold, deleting any one of the two second travel transfer plans; The above steps are repeated until the similarity of the feature data of any two second travel transfer plans is less than or equal to the threshold, and the first ordered sequence is constructed and output according to the remaining second travel transfer plans.

7. The method for rearranging travel transfer plans based on the two-stage method according to claim 2, characterized in that: The step of inputting the plurality of the first travel transfer solutions and the corresponding refined ranking scores into a preset sequence generation model to obtain a plurality of first ordered sequences comprises: Arranging a plurality of second travel transfer schemes and defining a window of a preset size, wherein the window includes a preset number of second travel transfer schemes; The window is placed at the starting position, and the window is slid to the right according to a preset step length until the window covers all the second trip transfer solutions; At each window position, a preset number of second travel transfer plans in the window are evaluated, wherein the evaluation includes calculating the total cost, the total duration and the comfort value; An optimal solution is determined according to the evaluation result, and a first ordered sequence is constructed and outputted according to a preset number of second travel transfer solutions corresponding to the optimal solution.

8. A system for rearranging travel transfer plans based on a two-stage method, characterized in that: It includes the sorting module, generation module, evaluation module and display module, among which: A refined ranking module, configured to obtain a plurality of first travel transfer schemes after refined ranking and corresponding refined ranking scores; A generation module configured to input the plurality of the first travel transfer schemes and the corresponding refined ranking scores into a preset sequence generation model to obtain a plurality of first ordered sequences, wherein the preset sequence generation model includes a plurality of generation algorithms, and each generation algorithm generates one or more first ordered sequences; An evaluation module configured to input a plurality of the first ordered sequences into a context-aware preset sequence evaluation model to determine a target ordered sequence with a highest exposure-to-purchase rate score; A display module is configured to display the target ordered sequence.

9. An electronic device, characterized in that: It includes a processor, a memory, a user interface and a network interface, the memory is used to store instructions, the user interface and the network interface are both used to communicate with other devices, and the processor is used to execute the instructions stored in the memory so that the electronic device executes the method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores instructions, and when the instructions are executed, the method according to any one of claims 1 to 7 is executed.

Images