Travel planning method and system based on big data prediction evaluation algorithm

Abstract

The invention discloses a travel planning method and system based on a big data prediction evaluation algorithm. The method comprises the steps of obtaining expected travel parameters of a user and constructing a network diagram; collecting travel data and evaluating all feasible routes according to the weight of the influence factor; generating a travel route scheme meeting user requirements according to an evaluation result; and predicting and calibrating the generated travel route scheme, and judging the feasibility of the travel route scheme to realize travel planning. A proper travel route planning scheme is obtained through mutual reference of the real-time environment parameters and the travel expected result, the weights of the influence factors of different users in different travel node processes are weighed, the influence factors of different travel nodes are assigned and calculated according to the weights, and the prediction accuracy of the travel route is improved.

Description

Technical Field

[0001] The present invention relates to the field of data prediction technology, and in particular to a travel planning method and system based on a big data prediction evaluation algorithm. Background Art

[0002] The current algorithms for recommending travel modes for tourists mainly include two categories. One is to manually formulate recommendation rules based on historical travel experience, but it is difficult to take into account the travel preferences of each user and the starting and ending points; the other is to use logistic regression models and other methods in combination with big data parameters to recommend travel modes. However, the travel mode ratio in this method is inaccurate to represent user travel preferences, resulting in the recommended travel mode not necessarily being the travel mode the user wants, and the recommendation accuracy and efficiency are poor. Therefore, a new travel route planning method is urgently needed.

[0003] Chinese patent document CN109918567A discloses a "travel mode recommendation method and device". The method adopts obtaining historical travel data; constructing a graph embedding model based on the historical travel data; the graph embedding model includes: user nodes, starting and ending nodes and travel mode nodes; the user nodes and the starting and ending nodes are respectively connected to the travel mode nodes through edges, and the weight of the edge is the frequency of the user node or the starting and ending nodes selecting the travel mode node; constructing a target optimization function based on the initial characterization vector of each node and the weight of each edge; adjusting the initial characterization vector of each node until the target function is less than a preset threshold, and obtaining the adjusted characterization vector of each node; recommending travel modes based on the adjusted characterization vector of each node. However, when planning the route, the patent did not take into account the impact of realistic influencing factors on travel route planning, resulting in inaccurate planning results. Summary of the invention

[0004] The present invention mainly solves the technical problem that the travel route planning accuracy of the original technical solution is not high, and provides a travel planning method and system based on big data prediction and evaluation algorithm. By weighing the weights of the influencing factors of different users in the travel route decision-making and personalized allocation, the corresponding required travel route plan is generated and the generated travel route is calibrated and optimized.

[0005] The above technical problem of the present invention is mainly solved by the following technical solution: The present invention comprises the following steps: S1 obtains the expected travel parameters of the user and constructs a network diagram; S2 collects travel data and evaluates all feasible routes according to the weights of influencing factors; S3 generates a travel route plan that meets user needs based on the evaluation results; S4 predicts and calibrates the generated travel route plan and determines the feasibility of the travel route plan to realize travel planning.

[0006] Preferably, the step S2 specifically includes: S21 lists and calculates the influencing factors between different travel nodes; S22 sets and assigns weights to different impact factors; S23 sets different weather conditions as penalty coefficients and obtains evaluation indexes of influencing factors in different schemes.

[0007] Preferably, the step S21 specifically includes listing and calculating the influencing factors between different travel nodes and setting the travel factors between the travel nodes set by the user as evaluation indicators, that is, evaluating and assigning relevant parameters between the nodes that affect the journey.

[0008] Preferably, the step S22 specifically includes obtaining the weights of different influencing factors by combining expert subjective assignment with data objective assignment through a combined weighting method of AHP hierarchical analysis method and CRITIC, wherein the AHP hierarchical analysis method for expert subjective judgment assignment specifically comprises: establishing an evaluation index system; then performing expert scoring by comparing the importance of the influencing factors; calculating the proportion of each influencing factor according to the scoring; and performing a consistency check on the calculated results.

[0009] As a preferred embodiment, the CRITIC objective assignment method is implemented as follows: setting n records and m indicators; calculating the standard deviation σ of the m indicators j And the correlation matrix R; calculate the information content contained in the m indicators and the weight of each indicator, combine the weights obtained by the AHP hierarchical analysis method and the weights obtained by the CRITIC objective assignment method to obtain the proportion of any influencing factor.

[0010] Preferably, the step S2 further comprises normalizing the influencing factor data to eliminate the singular sample data, wherein the normalized influencing factor comprises a positive feedback factor and a negative feedback factor, and the calculation formula for the positive feedback factor is as follows: Among them, x is the influencing factor, max(x) is the maximum value of the influencing factor, min(x) is the minimum value of the influencing factor, and Norm(x) is the normalized result of the influencing factor; The calculation formula for the negative feedback factor is as follows:

[0011] Preferably, a penalty coefficient based on different weather conditions is added to the influencing factor weight, and the influencing factor evaluation index of different schemes is calculated in combination with the influencing factor weight and the penalty coefficient, and the rationality of the route is adjusted according to the calculation result.

[0012] Preferably, the travel route calibration in step S4 includes a passenger flow prediction and evaluation algorithm, which counts the passenger flow going to the preset target node on the target day within n years and puts it into a first collection; counts the passenger flow on m days close to the target day each year and puts it into a second collection; calculates the passenger flow congestion index on m days close to the target day in previous years; and calculates the passenger flow prediction and evaluation index on the target day of the preset travel route in combination with the passenger flow congestion index.

[0013] Preferably, the passenger flow prediction and evaluation algorithm includes hypergraph prediction, which extracts the user's check-in behavior data to construct a historical travel sequence, constructs a corresponding hypergraph for each historical travel sequence, defines the travel route hypergraph, and classifies the nodes in the travel route into a unified set. Hyperedges are constructed from three perspectives: geographical distance, time series, and POI nodes. The hypergraph is constructed by splicing the hyperedge matrix to finally generate a prediction model for prediction.

[0014] A travel planning system based on a big data prediction and evaluation algorithm comprises: a data acquisition module, including a plurality of base station data acquisition and transmission devices, connected to a central processing module; the central processing module centrally processes travel parameter data to generate processing results; a real-time environmental parameter extraction module processes real-time environmental parameters and obtains extraction results from real-time environmental parameters; and a prediction module obtains a suitable travel route planning scheme by referring to real-time environmental parameters and expected travel results.

[0015] The beneficial effects of the present invention are as follows: the present invention weighs the weights of the influencing factors of different users in different travel nodes, assigns values ​​to the influencing factors of different travel nodes according to the weights, and calculates the penalty coefficients of different travel plans according to the assignment results to obtain several personalized travel plans that meet the needs of users. In addition, in the calculation process, the passenger flow prediction and evaluation algorithm for different target nodes in the previous n years and the hypergraph prediction are added to improve the accuracy of the travel plan calculation results, thereby improving the degree of fit between the personalized plan and the user's needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a flow chart of travel route planning of the present invention.

[0017] Figure 2 It is a structural diagram of travel route node selection of the present invention. DETAILED DESCRIPTION

[0018] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the technical solutions of the present application are further described in detail below through embodiments and in combination with the accompanying drawings. It should be understood that the specific implementation method described here is only an optimal embodiment of the present application, which is only used to explain the present application and does not limit the protection scope of the present application. All other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0019] Embodiment: A travel planning method and system based on a big data prediction and evaluation algorithm in this embodiment, such as Figure 1 and Figure 2 As shown, including the basic configuration of the system, firstly, the parameters are entered, and the basic parameters are configured in the system according to the user's travel intention. Including the user's planned travel time node and travel time range and the nodes along the travel route. Then the network diagram is constructed, the starting point and destination locations are obtained on the GIS map, and the starting point, waypoint and end point are used as nodes, and the lines between nodes are used as edges to establish a network diagram. Then data collection is carried out, and the collected data includes the following levels: 1. Ticket travel plan: Based on the big data analysis of ticketing systems, travel agencies, etc., the ticket price, travel time, travel time and class schedule of different travel modes based on the starting point to the end are collected; 2. Travel distance: Based on different travel modes, the travel distance under different plans from the starting point to the end point is obtained by statistics; 3. Play time: Get the best play time for different target locations; 4. Weather conditions: Collect the daily weather conditions during the travel time period.

[0020] After the preliminary data collection is completed, this technical solution also includes the calculation of the travel evaluation index between two points in the travel route, taking each point in the network diagram as the starting point or end point in turn, and enumerating the available ticketing schemes. For example, starting from a, when choosing the next destination, the available ticket information of ab, ac, and ad should be queried, and each scheme should be evaluated and the travel evaluation index calculated based on the possible influencing factor weights. The calculation of the above travel evaluation index includes the enumeration and calculation of influencing factors: the travel distance evaluation index from the starting point to the end point, denoted as D E ; The travel time evaluation index from the starting point to the end point is denoted by D T ; The travel cost evaluation index from the starting point to the end point is denoted by D C ; Transportation comfort evaluation index, denoted as D M (Expert subjective assignment method, 1-5 points, 5 is the highest); travel time suitability evaluation index, denoted as D S (Expert subjective assignment method, 1-5 points, 5 is the highest); terminal passenger flow prediction evaluation index, denoted as D F(Calculated based on the passenger flow prediction and evaluation algorithm). Then, data normalization is performed to normalize the data of similar influencing factors in different schemes. The calculation formula is as follows: Where x is the influencing factor, max(x) is the maximum value of the influencing factor, min(x) is the minimum value of the influencing factor, and Norm(x) is the normalized result of the influencing factor. In this embodiment, the four factors of travel distance, travel time, travel expenses, and passenger flow forecast evaluation are negative feedback indexes, which need to be processed using the reverse scoring conversion rule, such as corresponding a longer distance to a lower evaluation value. The specific method is as follows: It also includes the weight setting of influencing factors. Through the combined weighting method of AHP and CRITIC, the weight setting of different equipment types can better reflect the real scene situation by combining the subjective assignment of experts with the objective assignment of data, and improve the safety and rationality of scene risk assessment and prediction. The AHP hierarchical analysis method is used for expert subjective judgment assignment. The specific steps are as follows: 1. Establish an evaluation index system; 2. Expert scoring, by comparing the importance of two factors, the comparison matrix formula is as follows: 3. Calculate the weight of each factor; 4. Conduct consistency check on the obtained results; 5. Use the CRITIC objective assignment method to obtain weights.

[0021] The essence of the above assignment is to determine the objective weight based on the amount of data information, and use the standard deviation to characterize the comparison strength. The calculation steps and formula are as follows: 1. Suppose there are n records and m indicators; 2. Calculate the standard deviation σ of each indicator j and the correlation matrix R; 3. Calculate the amount of information contained in each indicator: 4. Calculate indicator weights: 5. For the i-th influencing factor, the weight obtained according to the AHP hierarchical analysis method Weights obtained by CRITIC objective evaluation method Calculate the comprehensive weight, the formula is as follows: In this technical solution, considering that weather conditions may affect the evaluation of the destination itinerary during the trip, a more accurate itinerary evaluation is obtained by defining weather conditions and their corresponding penalty coefficients. According to the expert resource acquisition questionnaire survey, the following weather conditions and their penalty coefficients θ are set: Sunny day: coefficient 1.0 (no penalty) Rainy day: coefficient 0.8 Snowy day: coefficient 0.6 Extreme high temperature (>35℃): coefficient 0.9 Extreme low temperature (<-10℃): The coefficient is 0.7. After completing the setting of the penalty coefficient, the evaluation index is calculated: If there are multiple solutions from starting point a to end point b, the impact factor evaluation index of the i-th solution is D E , D T , D C , D M , D S , D F , the corresponding weight is W E , W T , W C , W M , W S , W F , then the evaluation index of this scheme is calculated as follows: After calculation according to the above formula, the evaluation index is adjusted to filter out calculation results with large deviations and improve the accuracy of the calculation results.

[0022] After the evaluation index is adjusted in this technical solution, the overall travel route can be planned, including the time allocation for the destination. According to the travel time recommended by the system for each destination, the user's stay time at different destinations from the start date to the end date is automatically allocated. For example, in a 10-day trip, 1 day is allocated to place a, 3 days to place b, 2 days to place c, and 4 days to place d. Among them, the stay time at any place includes the travel time to reach this place by default, and the stay time at the last destination includes the return time by default.

[0023] Then calculate the travel evaluation index between two points. According to the travel plans queried from the ticketing system, calculate the travel evaluation index under different plans, and calculate the travel evaluation index of different plans under different destinations. Then calculate the comprehensive evaluation index of the entire route. Through the breadth-first algorithm, traverse the evaluation index of all travel plans from each departure point to the next destination to ensure that each destination is not repeated or missed. The comprehensive evaluation index of one of the travel routes is as follows: S=E a-b +E b-c +E c-d +E d-e Finally, based on the calculation results of the comprehensive evaluation index, the routes are sorted in order from large to small, and the top three routes are selected and recommended to users.

[0024] This technical solution also includes a passenger flow prediction and evaluation algorithm, which will predict passenger flow based on the target day (the date when the user arrives at this destination is defined as the target day). Therefore, different itinerary sequences will produce different passenger flow prediction results. This algorithm aims to predict the optimal route to ensure that the passenger flow prediction evaluation is optimal when the user arrives at each destination. At the same time, this algorithm takes into account the real-time nature of passenger flow and dynamically enters daily real-time data. Therefore, the closer the query is to the target day, the more accurate the result will be. Statistics on the passenger flow of the target day in each year in the past n years are recorded as set D n {f k-m ,…,f k-2 ,f k-1 ,f k}, where n represents 1, 2, 3, ..., f k Represents the target day passenger flow of the year. Count the target day passenger flow data for the past m days, recorded as a set where f k′ Indicates the passenger flow of the target day of the year. If the historical passenger flow data of a certain day already exists, it is directly entered into the set; if the historical passenger flow data of a certain day does not exist, the average passenger flow of the same day in the past five years is used as the predicted passenger flow of the day and entered into the set. Calculate the passenger flow congestion index of the target day in the past m days one by one and record it as Z, where is the sample mean, X i is the i-th sample data in the set. Calculate the average: Calculate the standard deviation: Calculate the passenger flow congestion index of the target day in previous years: Calculate the average passenger flow congestion index on target days over the years: Similarly, calculate the passenger flow congestion index Z′ on the target day of the year. Comprehensively calculate the passenger flow forecast evaluation index on the target day of the year: By weighing the weights of the influencing factors of different users at different travel nodes, assigning values ​​to the influencing factors of different travel nodes according to the weights, and calculating the penalty coefficients of different travel plans based on the assignment results, several personalized travel plans that meet the needs of users are obtained. In the calculation process, the passenger flow prediction evaluation algorithm and hypergraph prediction for different target nodes in the previous n years are added to improve the accuracy of the travel plan calculation results, thereby improving the degree of fit between personalized plans and user needs.

[0025] Example 2: In this example, the accuracy of the passenger flow prediction evaluation results of any travel node is improved by means of hypergraph prediction. In this example, the travel nodes of different users are defined as points of interest, i.e., POIs. Since the travel choices of users are derived from the influencing factors of user decision-making behavior, such as sequence effect, geographical effect, time effect, etc., they will affect the POI prediction recommendation results. Therefore, point of interest prediction recommendation is used to analyze the user's behavior and preferences, capture these influencing factors, predict and recommend the user's next interactive point of interest, and achieve effective recommendation. Hypergraph prediction requires the construction of several hypergraph structures. The hypergraph structure has significant advantages in expressing many-to-many data relationships through the characteristics of its hyperedges connecting any number of nodes. The learning ability of its neural network method is used to study the interactive relationship of the hypergraph structure.

[0026] This embodiment includes a technical solution for studying hypergraph structure using an unfolded neural network method. The unfolded hypergraph learning method is particularly targeted at decomposable hypergraphs, which decomposes complex hypergraph structures and converts them into traditional paired graph networks. By decomposing high-order interactions in the hypergraph, mature algorithms can be used for processing, and the correlation of complex passenger flow data at any travel node can be represented by a hypergraph structure. Finally, the neural network is generalized to effectively use the hypergraph structure to predict node passenger flow.

[0027] It also includes a technical solution for studying hypergraph structure using a non-expanded neural network method. The non-expanded hypergraph learning method is designed for processing indecomposable hypergraphs. This method avoids the decomposition of the hypergraph and directly models its inherent structure. In an indecomposable hypergraph, the nodes within the hyperedge often show close associations, however, any subset of the nodes within these hyperedges lacks this strong association. This learning method effectively captures the high-order dependencies between nodes by directly modeling the hypergraph structure. The self-attention-based hypergraph neural network deeply explores the complex structure and information in the hypergraph by introducing a self-attention mechanism, thereby achieving effective learning of non-uniform heterogeneous hypergraphs. This method provides a new perspective and effective tools for processing non-uniform heterogeneous hypergraphs. After the hypergraph is defined, the nodes included in it are placed into a unified set, namely the hyperedge set. For any hyperedge, it can contain more than two nodes. If all hyperedges in a hypergraph have the same size k, it is called a k-uniform hypergraph. Even if a hypergraph is k-uniform, it can still have different types of hyperedges because the types of nodes within the hyperedges can vary. The hyperedge prediction problem is defined as follows: the goal of the hyperedge prediction problem is to learn a function that meets the required conditions, so that the mutual relationships in the hypergraph structure that cannot be disassembled can be captured.

[0028] In the above-mentioned technical solutions for studying the hypergraph structure, there are multi-objective optimizations that are interrelated or conflicting. That is, different travel nodes are affected by time and space factors and passenger flow factors in the travel route plan, resulting in conflicting or interrelated problems, so that two nodes cannot be taken into account in the same travel route. For the multi-objective optimization problem, the problem is converted into MOP, and the specific formula is as follows: maxmizeF(x)=(f1(x), f2(x),…f n (x))T, subject to x∈Ω Where Ω is the decision space, F:Ω→Rm consists of m real-valued objective functions, Rm is the objective space, and {F(x)|x∈Ω} is the set of achievable objectives. If x∈Rn, all objectives are continuous, then the decision space Ω is defined as follows: Ω={x∈R n |hj(x)≤0,j=1,2,…m} Where hj(x) is a continuous function. Therefore, the multi-objective optimization problem is further solved by decomposing the multi-objective optimization problem into scalar optimization sub-problems through a decomposition-based multi-objective evolutionary algorithm. It solves these sub-problems simultaneously by evolving a group of solutions. In each generation, the population consists of the best solutions to each sub-problem. The neighborhood relationship between these sub-problems is defined based on the distance between their clustering coefficient vectors. The optimal solutions of two adjacent sub-problems should be similar. The scalar optimization problem is defined as follows, Where z * =(z *1 ,…,z *m )T is the reference point, that is, each point z *1 =max{fi(x)|x∈Ω}.

[0029] For each Pareto optimal point x*, there exists a weight vector λ such that the optimal solution of x* is the optimal solution in the formula. Different Pareto optimal solutions can be obtained by changing the weight vector. The neighborhood of the subproblem consists of all the subproblems whose weight vectors come from the neighborhood. Only the current solutions of the neighboring subproblems are used to optimize the subproblems. Assuming that the solutions on adjacent weights of different optimization objectives in the same travel node are similar, each weight has neighbors. After obtaining a new solution through the above method, it is necessary to replace the solution in the neighborhood. If the new solution generated is better than the solution obtained by the original individual, it can be replaced. The above steps are repeated so that all points in the population can be updated, and finally the solution of the multi-objective optimization problem converges to an approximate optimal solution.

[0030] Furthermore, after completing the multi-objective optimization on the same travel node, i.e., the point of interest, a point of interest prediction and recommendation model based on hypergraph learning is built on the same travel node. In this embodiment, a point of interest prediction and recommendation model based on hypergraph learning is established, which comprehensively considers key factors such as time, geography, and sequence, optimizes the historical travel route hypergraph construction process to capture the relevant influencing factors of point of interest prediction, and captures the high-order relationships in the user's historical check-in behavior through the learning of multiple parallel hypergraph attention convolutional networks. The model first constructs the user's historical travel route as a hypergraph, and then constructs specific sub-hypergraphs based on different intentions to capture the user's main intentions during the journey. Through the encoding of the sub-hypergraph, the user's core preferences under different intentions can be obtained, thereby achieving a more accurate prediction of the next point of interest. The model includes a given point of interest node set, where each element represents a tourist point of interest attribute vector, such as category, popularity, and geographic coordinates; a historical travel route sequence: a single historical travel route sequence is defined as ru, which represents the visit sequence of user u sorted by timestamp: the next point of interest prediction recommendation problem, given a point of interest set P and a historical travel sequence ru, the goal is to predict the user's next visited point of interest. The recommendation model will generate predicted preference scores for the POIs in the set P, and finally select the best several options for recommendation.

[0031] The specific execution steps include historical route hypergraph modeling. First, the user's check-in behavior data is extracted and constructed into a historical travel sequence. Given that user decision-making behavior is constrained by multiple factors, such as sequence effect, geographical effect and time effect, these factors significantly affect the prediction and recommendation results of POI (points of interest). In particular, considering that in these different perspectives, POIs present a many-to-many complex high-order relationship, this embodiment chooses to use a hypergraph model to simulate and construct these historical travel sequences. In order to accurately capture these complex relationships, a corresponding hypergraph is constructed for each historical travel sequence to achieve more accurate POI prediction and recommendation. The travel route hypergraph is defined as Gr, where the node set is the POI included in the historical travel route, and hyperedges are constructed from three perspectives: geographical distance, time series and the intention of POI nodes. In the process of splicing the hyperedge matrix to form a hypergraph, redundant hyperedges are deleted to generate a hyperedge set. The hyperedge set includes geographic hyperedges, and the geographical features of points of interest emphasized by location-based POI recommendations. In order to capture the impact of geographic distance information on POI prediction, POI nodes are divided by setting distance thresholds between POI nodes, and distance hyperedges are constructed for nodes within the same threshold interval to obtain distance hyperedge sets. Temporal hyperedges, in historical travel routes, the nearest travel nodes are usually more consistent with the current user's preferences, while the context contains the user's potential intentions. In order to accurately capture the time series information in historical travel routes. The weighted moving average method is applied to historical travel routes to effectively capture the temporal dynamics and trends in the data. By setting the value of the sliding window, the points of interest with a continuous visit length of the sliding window in the historical travel sequence are connected to form a temporal hyperedge.

[0032] Intent hyperedges,Intent factors play an important role in shaping users' historical travel behaviors, resulting in changes in perceived intentions for the same POI in different travel routes. Intent hyperedges are constructed to capture the correlation between POIs with specific intents. For different intents, an intent prototype is first constructed, the POI nodes are divided into k blocks, and each block is assigned to a specific user intent. The kth block of all POIs constitutes the kth intent space. By clustering the kth block of POIs in each intent subspace, an intent prototype is constructed, which represents the core intent of the corresponding subspace. For each POI node, its intent-aware embedding is initialized. Since the intent subspace contains the user's travel intention, the intent prototype is constructed by clustering all blocks in the kth order subspace. The intent hyperedge connects the first n POI nodes in the similarity value ranking, and the value of n affects the sparsity of the intent hyperedge. In the subsequent section, by discussing the value of n, we can find a suitable intent hyperedge to effectively capture the user's intent and assist in the prediction of the next interactive POI.

[0033] After completing the above steps, the next point of interest prediction recommendation is performed, and the historical travel route hypergraph containing multi-factor information is obtained through the hypergraph construction part. However, due to the check-in data of different travel intentions in the user's historical check-in data, the user's choice is often affected. In order to further capture the intentions contained in the POI nodes in the historical check-in data and solve the impact of the intention on the prediction results, this embodiment uses a k-parallel intention-aware encoder to parse the user's intentions. The original hypergraph is divided into multiple intention sub-hypergraphs. The specific intention hyperedge is combined with the geographic and time series hyperedges to form a specific intention sub-hypergraph. By propagating and learning specific blocks in the intention sub-hypergraph, the user's intention can be effectively captured and the mutual interference between different intentions can be reduced. In this way, the user's travel intention is accurately captured and the accuracy of predicting the probability of the next point is improved. Based on this, the passenger flow data of different travel nodes, i.e., points of interest in the past n years, can realize the passenger flow prediction evaluation of different target nodes, thereby improving the personalized satisfaction of formulating different travel routes for different users.

[0034] In this embodiment, the hypergraph is a generalization of a graph structure consisting of vertices and hyperedges, where the hyperedge can connect any number of vertices, surpassing the limitation that only two vertices are connected by one edge in the traditional graph. This feature of the hypergraph gives it a significant advantage in processing complex correlations in actual data, because the correlation of actual data often exceeds simple two-to-two relationships. The hypergraph structure can break through the limitations of linear space and map the complex relationships between data to POI prediction and recommendation based on hypergraph learning. Through the above method, when the user formulates a travel route method, a corresponding travel route hypergraph is constructed, considering multiple perspectives and integrating various auxiliary information; then the hypergraph is divided into specific sub-hypergraphs based on different intentions to facilitate embedding and propagation, capturing the main intentions and interest point tendencies of the user in the travel plan formulation, generating a representation of specific intentions, and realizing the prediction and recommendation of the next interest point. Finally, a multi-objective travel route planning model is established. By establishing a hybrid multi-objective optimization model, combining evolutionary algorithms and heuristic search, taking user experience as the core, and comprehensively considering multiple factors such as user preferences, travel time, cost, and satisfaction with attractions, tourists are provided with efficient and customized travel route planning.

[0035] The specific embodiments described herein are merely examples of the spirit of the present invention. The above embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be construed as limiting the scope of the present application. It should be noted that a technician in the technical field to which the present application belongs may make various modifications or supplements to the described specific embodiments or replace them in a similar manner, but will not deviate from the spirit of the present application or exceed the scope defined in the attached claims. For a person of ordinary skill in the art, multiple variations and improvements may also be made without departing from the concept of the present application. Therefore, the scope of protection of this patent shall be subject to the attached claims.

Claims

1. A travel planning method based on big data prediction and evaluation algorithm, characterized in that: The following steps are involved: S1 obtains the user's expected travel parameters and constructs a network graph; S2 collects travel data and evaluates all feasible routes according to the weights of influencing factors; S3 generates a travel route plan that meets user needs based on the evaluation results; S4 predicts and calibrates the generated travel route plan and determines the feasibility of the travel route plan to realize travel planning.

2. A travel planning method based on big data prediction and evaluation algorithm according to claim 1, characterized in that: The step S2 specifically includes: S21 lists and calculates the influencing factors between different travel nodes; S22 sets and assigns weights to different impact factors; S23 sets different weather conditions as penalty coefficients and obtains evaluation indexes of influencing factors in different schemes.

3. A travel planning method based on big data prediction and evaluation algorithm according to claim 2, characterized in that: The step S21 specifically includes enumerating and calculating the influencing factors between different travel nodes and setting the travel factors between the travel nodes set by the user as evaluation indicators, that is, evaluating and assigning relevant parameters affecting the travel between the nodes.

4. A travel planning method based on big data prediction and evaluation algorithm according to claim 2 or 3, characterized in that: The step S22 specifically includes, through the combined weighting method of AHP and CRITIC, obtaining the weights of different influencing factors by combining expert subjective assignment with data objective assignment, and the AHP hierarchical analysis method for expert subjective judgment assignment specifically includes: establishing an evaluation index system; then performing expert scoring by comparing the importance of the influencing factors; calculating the proportion of each influencing factor according to the scoring situation; and performing consistency check on the calculated results.

5. A travel planning method based on big data prediction and evaluation algorithm according to claim 4, characterized in that: The steps of the CRITIC objective value assignment method are as follows: set n records and m indicators; calculate the standard deviation σ of the m indicators j And the correlation matrix R; calculate the information content contained in the m indicators and the weight of each indicator, combine the weights obtained by the AHP hierarchical analysis method and the weights obtained by the CRITIC objective assignment method to obtain the proportion of any influencing factor.

6. A travel planning method based on big data prediction and evaluation algorithm according to claim 2 or 3, characterized in that: The step S2 also includes normalizing the impact factor data to eliminate the singular sample data. The normalized impact factor includes a positive feedback factor and a negative feedback factor. The calculation formula for the positive feedback factor is as follows: Among them, x is the influencing factor, max(x) is the maximum value of the influencing factor, min(x) is the minimum value of the influencing factor, and Norm(x) is the normalized result of the influencing factor; The calculation formula for the negative feedback factor is as follows:

7. A travel planning method based on big data prediction and evaluation algorithm according to claim 4, characterized in that: A penalty coefficient based on different weather conditions is added to the influencing factor weight, and the influencing factor evaluation index of different schemes is calculated in combination with the influencing factor weight and the penalty coefficient, and the rationality of the route is adjusted according to the calculation result.

8. A travel planning method based on big data prediction and evaluation algorithm according to claim 7, characterized in that: The travel route calibration in step S4 includes a passenger flow prediction and evaluation algorithm, which counts the passenger flow going to the preset target node on the target day within n years and puts it into a first collection; counts the passenger flow on m days close to the target day each year and puts it into a second collection; calculates the passenger flow congestion index on m days close to the target day in previous years; and calculates the passenger flow prediction and evaluation index on the target day of the preset travel route in combination with the passenger flow congestion index.

9. A travel planning method based on big data prediction and evaluation algorithm according to claim 8, characterized in that: The passenger flow prediction and evaluation algorithm includes hypergraph prediction, which extracts the user's check-in behavior data to construct a historical travel sequence, constructs a corresponding hypergraph for each historical travel sequence, defines the travel route hypergraph, and classifies the nodes in the travel route into a unified set. Hyperedges are constructed from three perspectives: geographical distance, time series, and POI nodes. The hypergraph is constructed by splicing the hyperedge matrix to finally generate a prediction model for prediction.

10. A travel planning system based on a big data prediction and evaluation algorithm, applicable to any one of the travel planning methods based on a big data prediction and evaluation algorithm in claims 1 to 9, characterized in that: It includes a connected data acquisition module and a central processing module, the data acquisition module includes several base station data acquisition and transmission devices, the central processing module is connected to a real-time environmental parameter extraction module and a prediction module, the real-time environmental parameter extraction module processes the real-time environmental parameters and obtains the extraction results in the real-time environmental parameters, and the prediction module obtains a suitable travel route planning plan by referring to the real-time environmental parameters and the expected travel results.

Images