Personalized itinerary recommendation method, system and device and storage medium

Abstract

The invention belongs to the field of itinerary planning, and particularly relates to a personalized itinerary recommendation method, system and device and a storage medium, an attention target concerned by a user during travel is extracted from big data, all POIs are divided into a plurality of refined categories, a prediction result is obtained according to travel data and a plurality of prediction models, and the prediction result is used for recommendation of a personalized itinerary. Performing personalized elimination and reservation on the POI data set according to a prediction result to obtain a preferred POI data set, generating an initialized travel in combination with travel preference obtained by prediction, then adopting crossover variation to obtain a plurality of to-be-selected travels, forming a to-be-screened recommended travel with the initialized travel, calculating fitness values of multiple types of targets of the to-be-screened recommended travel, and selecting the to-be-screened recommended travel according to the fitness values of multiple types of targets of the to-be-screened recommended travel. And further screening from the to-be-screened recommended itineraries according to the fitness value to obtain recommended itineraries, and then selecting the recommended itinerary with the highest preference probability from the recommended itineraries as the optimal recommended itinerary. And more concerned factors in the travel process of the user are considered, so that the recommended travel is closer to the demand of the user.

Description

Technical Field

[0001] The present invention belongs to the field of itinerary planning, and in particular relates to a personalized itinerary recommendation method, system, device and storage medium. Background Art

[0002] With the continuous improvement of economic level, people pay more and more attention to daily outings or travel to enrich their spiritual world. However, in an unfamiliar city, facing a large number of shopping malls, restaurants, rest places and many other different places, it is difficult to quickly plan a reasonable and effective travel route.

[0003] In related technologies, travel route planning is mainly single-objective optimization. After obtaining the travel locations that the user is interested in, a reasonable travel route is recommended based on the things that the user cares about most during the travel.

[0004] Regarding the above-mentioned related technologies, users usually care about many factors during their travels. If only a single factor is considered when making recommendations, the recommended itinerary may not be the travel route that the user is most interested in, which will bring a bad experience to the user. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide a personalized itinerary recommendation method, system, device and storage medium, taking into account more factors of concern in the user demand process, performing multi-objective optimization, and realizing personalized itinerary recommendation.

[0006] A personalized itinerary recommendation method, comprising:

[0007] Obtain user travel big data of user travel, travel data of current users to travel, several POIs and POI data corresponding to the POIs;

[0008] Classify several POIs to obtain POI refinement categories;

[0009] Analyze the user travel big data to extract the objects of interest that the user pays attention to when traveling;

[0010] According to the travel data of the current user to be traveled, the Ci prediction model, the Ti prediction model, the Wi prediction model and the Di prediction model, predict the number of overall POIs and the number of entertainment POIs that the user wants to include in the itinerary, the probability of the user selecting a certain POI refinement category, the probability of the user paying attention to the target of interest, and the probability of the user preferring the fine-grained attributes of the target of interest under a certain POI refinement category;

[0011] Screening POIs according to the preference probability and the focus target to obtain a candidate POI data set;

[0012] Design initialization itinerary;

[0013] According to the candidate POI data set, the total number of POIs, the number of entertainment POIs, the probability of the user selecting a certain POI refinement category, and the probability of the user paying attention to the attention target, the initialization itinerary is optimized through crossover mutation of a genetic algorithm to obtain several itineraries to be selected, and the fitness value of the initialization itinerary and the fitness value of the itinerary to be selected are calculated, and the fitness value is calculated through the attention target;

[0014] The initialized itinerary and the itinerary to be selected constitute a recommended itinerary to be screened;

[0015] Filter the recommended itinerary to be filtered according to the fitness value to obtain the recommended itinerary;

[0016] The best recommended itinerary for the user's travel is obtained based on the recommended itinerary and the user preference probability.

[0017] Optionally, the POI data includes basic attribute data and other attribute data, and the other attribute data includes grid attribute data, attribute data of the road network closest to a POI, popularity data, traffic accessibility data, POI detailed category data and POI cost data.

[0018] Optionally, the focus targets include traffic accessibility, regional vitality value, popularity, consumption range, travel distance, and POI detailed category preference.

[0019] Optionally, the prediction of the number of overall POIs that the user wants to include in the itinerary, the number of entertainment POIs, the probability of the user selecting a certain POI refinement category, the probability of the user paying attention to the target of interest, and the probability of the user's preference for fine-grained attributes of the target of interest under a certain POI refinement category based on the travel data of the current user to be traveling and the Ci prediction model, the Ti prediction model, the Wi prediction model, and the Di prediction model includes:

[0020] Acquire historical behavior data and user basic data of the current user to be on the road according to the travel data of the current user to be on the road;

[0021] Input the historical behavior data and user basic data into the Ci prediction model to predict the number of overall POIs and the number of entertainment POIs that the user wants to include in the itinerary;

[0022] Input the historical behavior data and user basic data into the Ti prediction model to predict the probability of the user selecting a certain POI refinement category;

[0023] Input the historical behavior data and user basic data into the Wi prediction model to predict the probability of the user paying attention to the target of interest;

[0024] The historical behavior data and user basic data are input into the Di prediction model to predict the user's preference probability for the fine-grained attribute preference of the target of interest under a certain POI refinement category.

[0025] Optionally, the performing personalized elimination and retention of POIs according to the preference probability and the focus target to obtain a candidate POI data set includes:

[0026] Constructing an expected scoring formula based on the user preference probability and the focus target;

[0027] The expected score formula is:

[0028] POI score=w1*p_score+w2*traf_score+w3*t_score+w4*c_score;

[0029] Among them, w1, w2, w3 and w4 are weights, p_score is popularity, traf_score is traffic accessibility score, t_score is POI detailed category preference score and c_score is consumption range score;

[0030] Scoring each POI according to the expected scoring formula to obtain a scoring result;

[0031] Based on the scoring results, POIs are eliminated and retained in a personalized manner to obtain a candidate POI data set.

[0032] Optionally, the design initialization process includes:

[0033] Get the number of initialized populations;

[0034] According to the total number of POIs, the number of the initialization population, the POI refinement category and the candidate POI data set, n initialization trips are randomly generated, where n is the number of the initialization population.

[0035] Optionally, the filtering the recommended itinerary to be filtered according to the fitness value to obtain the recommended itinerary includes:

[0036] Performing non-dominated sorting on the recommended itineraries to be screened according to the fitness values, and selecting the first m recommended itineraries to be screened as itineraries to be recommended;

[0037] Calculating the spatial distance between the recommended itinerary and a preset reference point;

[0038] The recommended itinerary is associated with the nearest reference point according to the spatial distance to obtain the number of associations of each reference point, and a recommended itinerary is obtained by filtering the recommended itineraries based on the number of associations.

[0039] A trip recommendation system, comprising:

[0040] An acquisition module is used to acquire user travel big data of user travel, travel data of current users to travel, several POIs and POI data corresponding to the POIs;

[0041] A classification module is used to classify several POIs and obtain POI detailed categories;

[0042] An extraction module, used to analyze the user travel big data and extract the objects of interest that the user pays attention to when traveling;

[0043] A prediction module, for predicting the number of overall POIs and entertainment POIs that the user wants to include in the itinerary, the probability of the user selecting a certain POI refinement category, the probability of the user paying attention to the target of interest, and the probability of the user preferring the fine-grained attributes of the target of interest under a certain POI refinement category, based on the travel data of the current user to be traveling, the Ci prediction model, the Ti prediction model, the Wi prediction model, and the Di prediction model;

[0044] A first screening module, configured to screen POIs according to the preference probability and the focus target to obtain a candidate POI data set;

[0045] Design module, used to design the initialization itinerary;

[0046] A crossover mutation module is used to optimize the initialization trip through crossover mutation of a genetic algorithm according to the candidate POI data set, the total number of POIs, the number of entertainment POIs, the probability of the user selecting a certain POI refinement category, and the probability of the user paying attention to the attention target, to obtain several trips to be selected, and calculate the fitness value of the initialization trip and the fitness value of the trip to be selected, wherein the fitness value is calculated through the attention target;

[0047] The initialized itinerary and the itinerary to be selected constitute a recommended itinerary to be screened;

[0048] A second screening module, configured to screen the recommended itineraries to be screened according to the fitness value to obtain the recommended itineraries;

[0049] The recommendation module is used to obtain the best recommended itinerary for the user's travel based on the recommended itinerary and the user preference probability.

[0050] A terminal device includes a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and when the processor loads and executes the computer program, the above-mentioned personalized itinerary recommendation method is adopted.

[0051] A computer-readable storage medium stores a computer program, and when the computer program is loaded and executed by a processor, the above-mentioned personalized itinerary recommendation method is adopted.

[0052] The beneficial effects of the present invention are as follows: by analyzing the user travel big data, the focus targets that the user pays attention to when traveling are extracted, all POIs are divided into several categories, and according to the travel data of the current user to be traveled and multiple prediction models, the total number of POIs, the number of entertainment POIs, the probability of the user selecting a certain POI refinement category, the user's focus probability on the focus target, and the user's preference probability for the fine-grained attribute preference of the focus target under a certain POI refinement category are predicted when the user to be traveled again. Then, according to the preference probability and the focus target, the POIs that the current user to be traveled may choose are first screened to obtain a candidate POI data set, and then the prediction results of multiple prediction models and the candidate POI data set are used to generate an initialization itinerary, and then cross-mutation is used to obtain several itineraries to be selected, and the itinerary to be selected and the initialization itinerary constitute a recommended itinerary to be screened, and the fitness value of the recommended itinerary to be screened is calculated. According to the fitness value, the recommended itinerary to be screened is further screened to obtain a recommended itinerary, and then the one with the highest preference probability is selected from the recommended itinerary as the recommended itinerary. Considering the factors that the user pays more attention to during the travel process, the recommended personalized itinerary is closer to the needs of the user. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 A schematic diagram of a process of a personalized itinerary recommendation method of the present invention;

[0054] Figure 2 A model parameter table of a Ci prediction model, a Ti prediction model, a Wi prediction model, and a Di prediction model of a personalized itinerary recommendation method of the present invention;

[0055] Figure 3 A schematic diagram of population selection for a personalized itinerary recommendation method of the present invention;

[0056] Figure 4 A schematic diagram of recommendation results of a personalized itinerary recommendation method according to the present invention. DETAILED DESCRIPTION

[0057] A personalized itinerary recommendation method, such as Figure 1 As shown, the present invention includes:

[0058] S100, obtaining user travel big data of user travel, travel data of current users to travel, a number of POIs and POI data corresponding to the POIs.

[0059] Specifically, user travel big data mainly refers to the data obtained from the Internet on the destinations that several users pay attention to during their travels, spending data, popularity of POIs in a region, etc. The travel data of users who are currently preparing to travel refers to the data of users who are preparing to make itinerary recommendations, which can be obtained through questionnaires, including the detailed categories of POIs selected by users during their previous travels, the number of POIs, the spending range, and the destinations that they care about when traveling.

[0060] POI refers to a point of interest for travel, that is, a possible destination, such as a certain shopping mall. POI data is data related to POI in a certain area. POI data includes basic attribute data and other attribute data. Other attribute data includes grid attribute data, attribute data of the road network closest to a certain POI, popularity data, traffic accessibility data, POI detailed category data, and POI cost data.

[0061] POI data includes basic attribute data and other attribute data. Other attribute data include grid attribute data, attribute data of the road network closest to a POI, popularity data, traffic accessibility data, POI detailed category data and POI cost data.

[0062] Grid attribute data, which represents the grid where the POI is located, and is used to assign regional vitality values ​​to the POI.

[0063] The nearest road network attribute data. This data represents the nearest road network code (RID) to the POI, which is used to assign road accessibility to the POI.

[0064] Popularity. This attribute is the normalized popularity score of the POI.

[0065] Traffic accessibility: This attribute data is the traffic accessibility of the road closest to the POI and has been normalized.

[0066] POI detailed category data includes categories and detailed categories. This attribute data is a detailed classification of various POIs, that is, each major category is further divided into multiple minor categories.

[0067] POI spending data. This attribute data is the average per capita spending of the POI.

[0068] S110: Classify a number of POIs to obtain POI refinement categories.

[0069] Specifically, based on the user travel big data of users, all POIs are divided into food, shopping, sports, social, leisure and scenery categories.

[0070] S120: Analyze the user travel big data to extract the targets that the user pays attention to when traveling.

[0071] Specifically, the focus of users when traveling refers to the aspects of experience that users pay more attention to when traveling, including traffic accessibility, regional vitality value, popularity, consumption range, travel distance, and POI detailed category preferences. User travel big data refers to the collection of data on the targets that users pay attention to when other users traveled in the past.

[0072] Among them, traffic accessibility refers to the fact that users are more willing to choose places with convenient transportation during their travel. Here, the travel degree is used to measure the traffic convenience of the road network. The calculation formula of the travel degree is as follows:

[0073]

[0074] Among them, f1 represents the fitness function value, c represents the number of POIs, TPBt i It represents the travel degree of a street. The travel degree (TPBt) calculates the probability or number of the shortest paths from all spaces to all other spaces of a street segment, which can reflect the potential of people's travel activities in the road network.

[0075] The final calculation result of the travel degree is reflected in the road network. We project the POI to the nearest road network to obtain the identification (RID) and corresponding travel degree (TwoPhaseBetweenness, TPBt) of the road network as the traffic accessibility attribute of the POI. Due to the large difference in TPBt scores, in order to reduce the error, we logarithmically normalize the TPBt attribute value of each POI.

[0076] The traffic accessibility fitness of a trip is the sum of the TPBt scores of all POIs.

[0077]

[0078] Among them, OD(y,z,x) is the assignment rule of the shortest path under different conditions, searching for the shortest topological path between nodes y and z passing through node x within the search radius R, TPBt(x) is the travel degree of node x; Links(y) is the total number of nodes within the search radius R for each node y. In discrete space, if the point is within the search radius, the P(y) value is 1, otherwise it is 0; in continuous space, the P(y) value depends on the ratio of the radius to the length of the line segment, which is between 0 and 1, N is the set of all road segments in the study area, R yis the set of other road segments that can be reached within a given radius R starting from road segment y.

[0079] The shortest topological path is calculated as follows:

[0080]

[0081] The consumption range refers to the price range that users have psychological expectations of when choosing a POI, so it is necessary to make the consumption range as close to the user's preference as possible. Therefore, we have classified consumption for various types of POIs, and the classification standard refers to the consumption range of Dianping.com. In the model prediction part, the user's consumption range for a specific type of POI is also used as a prediction attribute, and the segmented score is given according to the predicted preference range. The more likely the user is to prefer this consumption range, the higher the POI score for this consumption range. The cost target of the itinerary is the sum of the cost scores of each POI, and the target calculation formula is as follows:

[0082]

[0083] Among them, cost i represents the cost of the i-th scenic spot, ω represents the score of the cost given by the mapping function, and f2 represents the total score of c scenic spots.

[0084] Popularity refers to the popularity of a POI, that is, its popularity. When choosing a POI, some users prefer POIs with higher popularity. The popularity rankings of various POIs are graded and normalized to reduce errors. The traffic accessibility target of the trip is the sum of the popularity scores of each POI. The target calculation formula is as follows:

[0085]

[0086] Among them, popu i is the popularity score of the i-th attraction, and β represents the score of the cost given by the mapping function.

[0087] Trip distance refers to the distance cost that users must keep within a certain range when traveling. Since the itinerary needs to be dynamically optimized to approach the user's target needs and improve the degree of personalization, the location of the itinerary POI points is also constantly changing. The trip distance is chosen as the fitness in the optimization process to constrain the trip distance. In order to find the global distance optimal solution for uncertain paths, the simulated annealing algorithm SA is used. SA is a random optimization algorithm that is better than the general greedy algorithm because it will accept a solution that is worse than the current solution with a certain probability, which makes the distance calculation more likely to jump out of the local optimal solution to find the global distance optimal solution. The calculation formula for the trip distance target is as follows:

[0088]

[0089] Among them, s i represents the longitude and latitude coordinates of the starting point, and p ij are the longitude and latitude coordinates of the candidate point. h(s i , p ij ) is the distance between two points calculated by the haversine formula. The calculation order of the points in the formula is determined by the simulated annealing algorithm.

[0090] The calculation formula of SA is as follows:

[0091]

[0092] Assume that the previous state of the POI sequence is x(n). Through a certain rule, the system changes to x(n + 1). Correspondingly, the energy of the system changes from E(n) to E(n + 1). The acceptance probability P of the system changing from x(n) to x(n + 1) can be expressed by the calculation formula of SA. The internal energy E in the formula is simulated as the objective function value f4 of the distance, and T represents the temperature at this time, which is defined in the initial parameters and decreases with the iteration of SA.

[0093] If the energy decreases, then this transfer is accepted (with a probability of 1). If the energy increases, it means that the system is farther away from the global optimal position. However, at this time, the algorithm will not immediately discard it, but perform a probability operation: generate a uniformly distributed random number ∈ in the interval (0, 1). If ∈ < P, then this transfer is accepted; otherwise, the transfer is rejected and the next step is entered, and the cycle repeats. Among them, the probability P is determined by the change in energy and T, and the probability of P changes dynamically.

[0094] Ultimately, what SA obtains is the shortest Euclidean distance of the POI itinerary without a definite starting point. The starting point of the itinerary is not a fixed point, but a point that can be changed by the return of the next iteration. Through this distance, it can be ensured that the path is always within a reasonable range during the optimization process.

[0095] f4 is realized by randomly defining the starting point in the POI sequence, calculating the Haversine distance from the previous point to the next point, and accumulating it as the path distance. Haversine calculates the distance between two points on the earth through longitude and latitude. The Haversine formula is as follows:

[0096]

[0097] The regional vitality value comprehensively reflects the number of candidate POIs in the area where the user to travel is active, as well as the richness of the POI categories, which can give users a richer and more flexible choice, so that the recommended itinerary is more likely to satisfy users in actual applications. Considering the number of POIs and the diversity of POIs in the grid of the area where the POI is located, it is ensured that the geographical location of the POI is relatively excellent when recommending the itinerary.

[0098] The vitality value feature of the grid is combined with the POI diversity and the number of POIs, and is assigned to the POIs in the area as a new attribute of the POI. The vitality value of the trip is the combination of the vitality values ​​of the POIs in the trip. The fitness target f5 is expressed as follows:

[0099]

[0100] N represents the number of POIs in the grid where the POI is located, and SDI represents the mixed degree of the grid where the POI is located.

[0101] The weight of the number of POIs and the diversity of POIs in the regional vitality value is designed to be 6:4, which is provided by the questionnaire survey of this method. First, the region is divided into fishing nets, the correspondence between grids and POIs is defined, and the Shannon diversity and the number of POIs in each grid area are calculated. Shannon entropy is a more classic method for calculating the degree of mixing, and the diversity is calculated as follows:

[0102]

[0103] p i is the area ratio of POI of type i within the perception range of the grid, m is the total number of POI detailed categories in this range, when there is only one type, the Shannon diversity value is 0, and when the proportions of each type are equal, the entropy value is the largest.

[0104] POI refinement category preference refers to the POIs that users are more interested in during their trip. In order to ensure the personalization of itinerary planning, after deleting the POIs with poor performance, the method retains the POI category preference score as the sixth objective function. Through this objective function, it can be ensured that in the final recommendation sequence, the POI refinement category is closer to user needs. The POI category preference score is calculated as follows:

[0105]

[0106] S130. Based on the travel data of the current user to be traveling, the Ci prediction model, the Ti prediction model, the Wi prediction model and the Di prediction model, predict the total number of POIs that the user hopes to include in the itinerary, the number of entertainment POIs, the probability of the user selecting a certain POI refinement category, the probability of the user paying attention to the target of interest, and the probability of the user's preference for the fine-grained attributes of the target of interest under a certain POI refinement category.

[0107] Specifically include:

[0108] S1300. Obtain historical behavior data and user basic data of the current user to be on the road according to travel data of the current user to be on the road.

[0109] Specifically, historical behavior data includes data on the detailed categories of POIs selected by users in past trips and the number of POIs selected during each trip, while basic user data includes user gender, age, occupation, etc.

[0110] S1301. Input historical behavior data and user basic data into the Ci prediction model to predict the number of overall POIs and the number of entertainment POIs that the user hopes to include in the itinerary.

[0111] S1302: Input the historical behavior data and the basic user data into the Ti prediction model to predict the probability of the user selecting a certain POI refinement category.

[0112] S1303: Input the historical behavior data and the user basic data into the Wi prediction model to predict the probability of the user paying attention to the target.

[0113] S1304: Input the historical behavior data and the user basic data into the Di prediction model to predict the user's preference probability for the fine-grained attribute preference of the target of interest under a certain POI refinement category.

[0114] Specifically, according to the user survey analysis, various POI fine-grained attributes that users care about are selected to build model predictions. The models are divided into cold start series and full attribute prediction series. The ultimate goal is to obtain user fine-grained attribute preference data. The user's complete input attribute data is represented by Ui = {Bi, Hi, Wi}, where Bi represents the user's basic attribute data; Hi represents the user's historical behavior selection data; Wi represents the user's historical preference target data. The model input data is reorganized by Ui and varies from model to model. The input attribute data of the cold start series model is Ui = {Bi}, and other related attributes are generated by the model hierarchically and taken into account in the next level of prediction. The complete user preference attribute data output by the model is represented by Pi = {Ci, Ti, Wi, Di}, where Ci and Ti attributes are obtained by disassembling Hi, Wi represents the user's desired preference target data, and Di represents the user's detailed feature preference for various POI data.

[0115] Model parameters such as Figure 2 shown.

[0116] Specifically, under both series, the models can be divided into (1) Ci prediction model, which obtains the total number of POIs and entertainment POIs that users want to include in their trip; (2) Ti prediction model, which obtains the detailed categories of POIs that users may want to choose; (3) Wi prediction model, which understands which conditions users want to have better in their trip, that is, which goals users are more looking forward to; (4) Di prediction model, under which POI attributes and user needs are further refined. The model ultimately obtains the user's detailed feature preferences for various POI data.

[0117] The model uses a multi-layer feedforward fully connected neural network optimization process, selects loss functions suitable for each model, and takes into account the idea of ​​multi-task learning in model design, sharing input layers and some hidden layers. Model types include multi-task multi-classification neural network model, multi-label classification neural network model, etc.

[0118] In general, the implementation logic of the user preference prediction model is to input user attribute data and obtain (1) Ci in the recommended itinerary. Infer the number of POIs that the user wants to include in the itinerary; (2) Ti in the recommended itinerary. Infer which types of POIs the user is more likely to choose under the quantity constraint; (3) Wi in the recommended itinerary. Infer the degree to which the user cares about multiple goals in the itinerary planning problem; (4) Di in the recommended itinerary. Infer which POIs the user prefers when selecting POIs, or which POIs will score higher in the eyes of the user.

[0119] S140 , performing personalized elimination and retention of POIs according to preference probabilities and focus targets to obtain a candidate POI data set.

[0120] Specifically, user preference probability describes the degree to which users care about the target under a certain POI refinement category. Therefore, POIs with low user preference probability can be deleted from all POIs through user preference probability optimization. Specifically, according to the six major focus targets, the relevant POI attributes that may be used are quantified. According to user preferences, personalized scoring is achieved for multiple POI categories, and POIs with poor performance are eliminated. The POI scoring content is divided into three modules: (1) basic content scoring; (2) preference-related attribute scoring; (3) target expectation scoring. The basic content scoring scores the popularity and traffic accessibility attributes of the POI. The higher the popularity and traffic accessibility attribute values, the better. The preference-related attribute scoring scores the sub-category and price range of the POI. The sub-category scoring scores the user's preference for the POI of this category. For the price attribute, the article maps the price range according to the commonly used classification in daily travel. The more it conforms to the user's preferred consumption range, the higher the score. The target expectation score comes from the prediction of the expectation level of multiple scoring targets. This level will be used as the scoring weight to calculate the total score of the POI.

[0121] The total POI score is calculated as follows:

[0122] POI score=w1*p_score+w2*traf_score+w3*t_score+w4*c_score.

[0123] Among them, w1, w2, w3 and w4 are weights, p_score is popularity, traf_score is traffic accessibility score, t_score is POI detailed category preference score and c_score is consumption range score.

[0124] The popularity score and traffic accessibility score can be assigned hierarchically according to the public ranking. For example, the top 20 in the area are assigned a score of 1, the top 50 are assigned a score of 0.8, and the top 100 are assigned a score of 0.6. The POI detailed category preference score and the c_score consumption range score are mapped through a mapping function. The user's highest probability of preference for the fine-grained attributes of the target of interest under a certain POI detailed category is assigned a score of 1, the second probability is 0.8, and so on, which are divided into five mapping intervals.

[0125] S150, designing an initialization itinerary.

[0126] Specifically include:

[0127] S1500, obtaining the number of initialized populations.

[0128] Specifically, the number of the initialization population is the number of initialization trips to be generated.

[0129] S1501. Randomly generate n initialization trips according to the total number of POIs, the number of initialization populations, POI refinement categories, and candidate POI data sets, where n is the number of initialization populations.

[0130] Specifically, when designing the initialization itinerary, the gene library is the screened POI, the total number of itinerary POIs is obtained by the Ci prediction model, the Ti prediction model predicts the set of POI refinement categories available for selection, and the itinerary is designed as a chromosome, which is encoded as Ii={p1,p2…,pc}, where i represents the recommended i-th itinerary, gene pj represents the j-th recommended POI in the itinerary, c is the total number of itinerary POIs, and then n chromosomes are randomly generated as the initialization itinerary.

[0131] S160, optimizing the initialization itinerary through crossover mutation of a genetic algorithm according to the candidate POI data set, the total number of POIs, the number of entertainment POIs, the probability of a user selecting a certain POI refinement category, and the probability of a user paying attention to a focus target, obtaining several itineraries to be selected, and calculating the fitness value of the initialization itinerary and the fitness value of the itinerary to be selected, wherein the fitness value is calculated based on the focus target.

[0132] Specifically, after generating the initialized population, the crossover and mutation operators in genetics are used to generate offspring. The purpose of the crossover operator is to generate offspring chromosomes that inherit the results of the parent chromosomes; the purpose of the mutation operator is to generate offspring chromosomes by changing random genes. Subsequently, the fitness calculation is performed based on the six major goals, namely the focus goals, and the goals are optimized as the fitness in the optimization process. In this work, six fitness values ​​corresponding to the goals were calculated: {-f1, -f2, -f3, f4, -f5, -f6}. It is worth noting that when performing crossover mutation, the present application is different from conventional crossover mutation. When performing crossover mutation, the categories of each gene position in the chromosome are fixed, and the crossover mutation realizes positioning crossover and fixed-point mutation according to the gene site to ensure the continuation of multiple categories of genes in the chromosome.

[0133] S170: The initialized itinerary and the itinerary to be selected constitute the recommended itinerary to be screened.

[0134] S180 , screening the recommended itineraries to be screened according to the fitness values ​​to obtain recommended itineraries.

[0135] The recommended itineraries to be screened are screened according to the fitness value, and the recommended itineraries include:

[0136] S1800 , performing non-dominated sorting on the recommended itineraries to be screened according to the fitness values, and selecting the first m recommended itineraries to be screened as the itineraries to be recommended.

[0137] Specifically, first calculate the fitness value of the target in the recommended itinerary, and select the parent (Pt) and child (Ot) that have completed the fitness calculation. The selection of these populations is performed by non-dominated quick sorting and niche-preserving selection. According to the fitness value, the individuals are quickly divided into multiple Pareto fronts, so as to perform non-dominated sorting and niche-preserving selection. Figure 3 As shown in the figure, F1 dominates all other individuals, that is, it performs better than individuals of other levels in any fitness. F2 individuals are only dominated by F1 and dominate F3 and later levels. The higher the level, the better the individual is, and the easier it is to be selected into the next generation (Pt).

[0138] S1801. Calculate the spatial distance between the recommended itinerary and a preset reference point.

[0139] S1802: Associating the itinerary to be recommended with the nearest reference point according to the spatial distance, obtaining the number of associations of each reference point, and filtering the recommended itineraries based on the number of associations to obtain the recommended itineraries.

[0140] The niche-preserving selection module maintains the diversity of the population by introducing widely distributed reference points. When Ot selects the last dominance level, it only selects part of it. In this process, the NSGA-III algorithm is first used to generate hyperplanes of two reference points in the inner and outer layers, and two layers of reference points are generated that are evenly distributed in the solution space according to the target dimension and population size. Then, the minimum values ​​of the six objective functions are used as ideal points, and the target fitness of the individuals in the population is normalized according to the ideal points and extreme points. Third, find the reference line closest to each individual (the line connecting the reference point and the ideal point), connect the individual with the reference line, and finally associate the individual with the corresponding reference point. Individuals corresponding to reference points with fewer connections should be retained, which is conducive to maintaining the diversity of the solution set. Next, the above process will be iterated until the itinerary meets the number of iterations required.

[0141] S190: Obtain the best recommended itinerary based on the recommended itinerary and the preference probability.

[0142] Specifically, the recommended itinerary includes several travel routes, and then according to the user's preference probability, the result population with the maximum probability preference is selected for recommendation.

[0143] The POIs of the recommended results are projected to the nearest local road network structure and the final path is generated. The recommended results will be presented to the user in the form of a path, such as Figure 4 shown.

[0144] A trip recommendation system, comprising:

[0145] An acquisition module is used to acquire user travel big data of user travel, travel data of current users to travel, several POIs and POI data corresponding to the POIs;

[0146] A classification module is used to classify several POIs and obtain POI detailed categories;

[0147] The extraction module is used to analyze the user travel big data and extract the objects of interest that the user pays attention to when traveling;

[0148] A prediction module is used to predict the total number of POIs that the user wants to include in the itinerary, the number of entertainment POIs, the probability of the user selecting a certain POI refinement category, the probability of the user paying attention to the target of interest, and the probability of the user preferring the fine-grained attributes of the target of interest under a certain POI refinement category based on the travel data of the current user to be traveling, the Ci prediction model, the Ti prediction model, the Wi prediction model, and the Di prediction model;

[0149] The first screening module is used to screen POIs according to preference probability and focus targets to obtain a candidate POI data set;

[0150] Design module, used to design the initialization itinerary;

[0151] The crossover mutation module is used to optimize the initialization trip through the crossover mutation of the genetic algorithm according to the candidate POI data set, the total number of POIs, the number of entertainment POIs, the probability of users selecting a POI refinement category, and the probability of users paying attention to the attention target, and obtain several trips to be selected, and calculate the fitness value of the initialization trip and the fitness value of the trip to be selected. The fitness value is calculated by paying attention to the target;

[0152] The initialization itinerary and the itinerary to be selected constitute the recommended itinerary to be screened;

[0153] The second screening module is used to screen the recommended itineraries to be screened according to the fitness value to obtain the recommended itineraries;

[0154] The recommendation module is used to obtain the best recommended itinerary for the user based on the recommended itinerary and the user's preference probability.

[0155] An embodiment of the present application also discloses a terminal device, including a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and when the processor loads and executes the computer program, a personalized itinerary recommendation method, system, device and storage medium are adopted.

[0156] The terminal device may be a computer device such as a desktop computer, a laptop computer or a cloud server, and the terminal device includes but is not limited to a processor and a memory. For example, the terminal device may also include input and output devices, a network access device and a bus.

[0157] Among them, the processor can adopt a central processing unit (CPU). Of course, according to actual usage, other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. can also be adopted. The general-purpose processor can adopt a microprocessor or any conventional processor, etc., and this application does not impose any restrictions on this.

[0158] Among them, the memory can be an internal storage unit of the terminal device, such as a hard disk or memory of the terminal device, or it can be an external storage device of the terminal device, such as a plug-in hard disk, a smart memory card (SMC), a secure digital card (SD) or a flash memory card (FC) equipped on the terminal device, etc., and the memory can also be a combination of an internal storage unit and an external storage device of the terminal device. The memory is used to store computer programs and other programs and data required by the terminal device. The memory can also be used to temporarily store data that has been output or is to be output, and this application does not impose any restrictions on this.

[0159] Among them, through this terminal device, a personalized itinerary recommendation method, system, device and storage medium in the above-mentioned embodiment are stored in the memory of the terminal device, and are loaded and executed on the processor of the terminal device for easy use.

[0160] An embodiment of the present application further discloses a computer-readable storage medium, and the computer-readable storage medium stores a computer program, wherein when the computer program is executed by a processor, a personalized itinerary recommendation method, system, device and storage medium in the above-mentioned embodiment is adopted.

[0161] Among them, the computer program can be stored in a computer-readable medium, the computer program includes computer program code, the computer program code can be in the form of source code, object code, executable file or certain middleware, etc. The computer-readable medium includes any entity or device that can carry the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the computer-readable medium includes but is not limited to the above-mentioned components.

[0162] Among them, through this computer-readable storage medium, a personalized itinerary recommendation method, system, device and storage medium in the above-mentioned embodiment are stored in a computer-readable storage medium, and are loaded and executed on a processor to facilitate the storage and application of the above-mentioned method.

[0163] A person skilled in the art should understand that the discussion of any of the above embodiments is merely illustrative and is not intended to imply that the scope of protection of the present application is limited to these examples. In line with the concept of the present application, the technical features in the above embodiments or different embodiments may be combined, the steps may be implemented in any order, and there are many other variations of different aspects of one or more embodiments of the present application as above, which are not provided in detail for the sake of simplicity.

[0164] One or more embodiments of the present application are intended to cover all such substitutions, modifications and variations that fall within the broad scope of the present application. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of one or more embodiments of the present application should be included in the protection scope of the present application.

Claims

1. A personalized itinerary recommendation method, characterized in that: include: Obtain user travel big data of user travel, travel data of current users to travel, several POIs and POI data corresponding to the POIs; Classify several POIs to obtain POI refinement categories; Analyze the user travel big data to extract the objects of interest that the user pays attention to when traveling; According to the travel data of the current user to be traveled, the Ci prediction model, the Ti prediction model, the Wi prediction model and the Di prediction model, predict the number of overall POIs and the number of entertainment POIs that the user wants to include in the itinerary, the probability of the user selecting a certain POI refinement category, the probability of the user paying attention to the target of interest, and the probability of the user preferring the fine-grained attributes of the target of interest under a certain POI refinement category; Perform personalized elimination and retention of POIs according to the preference probability and the focus target to obtain a candidate POI data set; Design initialization itinerary; According to the candidate POI data set, the total number of POIs, the number of entertainment POIs, the probability of the user selecting a certain POI refinement category, and the probability of the user paying attention to the attention target, the initialization itinerary is optimized through crossover mutation of a genetic algorithm to obtain several itineraries to be selected, and the fitness value of the initialization itinerary and the fitness value of the itinerary to be selected are calculated, and the fitness value is calculated through the attention target; The initialized itinerary and the itinerary to be selected constitute a recommended itinerary to be screened; Filter the recommended itinerary to be filtered according to the fitness value to obtain the recommended itinerary; The best recommended itinerary for the user's travel is obtained based on the recommended itinerary and the preference probability.

2. The personalized itinerary recommendation method according to claim 1, characterized in that: The POI data includes basic attribute data and other attribute data, wherein the other attribute data includes grid attribute data, attribute data of the road network closest to a POI, popularity data, traffic accessibility data, POI detailed category data and POI cost data.

3. A personalized itinerary recommendation method as claimed in claim 1, characterized in that: The focus targets include traffic accessibility, regional vitality value, popularity, consumption range, travel distance and POI detailed category preference.

4. The personalized itinerary recommendation method according to claim 1, characterized in that: The method of predicting the number of overall POIs and entertainment POIs that the user wants to include in the itinerary, the probability of the user selecting a certain POI refinement category, the probability of the user paying attention to the target of interest, and the probability of the user's preference for fine-grained attributes of the target of interest under a certain POI refinement category based on the travel data of the current user to be traveling and the Ci prediction model, the Ti prediction model, the Wi prediction model, and the Di prediction model includes: Acquire historical behavior data and user basic data of the current user to be on the road according to the travel data of the current user to be on the road; Input the historical behavior data and user basic data into the Ci prediction model to predict the number of overall POIs and the number of entertainment POIs that the user wants to include in the itinerary; Input the historical behavior data and user basic data into the Ti prediction model to predict the probability of the user selecting a certain POI refinement category; Input the historical behavior data and user basic data into the Wi prediction model to predict the probability of the user paying attention to the target of interest; The historical behavior data and user basic data are input into the Di prediction model to predict the user's preference probability for the fine-grained attribute preference of the target of interest under a certain POI refinement category.

5. The personalized itinerary recommendation method according to claim 1, characterized in that: The POIs are eliminated and retained according to the preference probability and the focus target to obtain a candidate POI data set, which includes: Constructing an expected scoring formula based on the user preference probability and the focus target; The expected score formula is: POI score=w1*p_score+w2*traf_score+w3*t_score+w4*c_score; Among them, w1, w2, w3 and w4 are weights, p_score is popularity, traf_score is traffic accessibility score, t_score is POI detailed category preference score and c_score is consumption range score; Scoring each POI according to the expected scoring formula to obtain a scoring result; Based on the scoring results, POIs are eliminated and retained in a personalized manner to obtain a candidate POI data set.

6. The personalized itinerary recommendation method according to claim 1, characterized in that: The design initialization process includes: Get the number of initialized populations; According to the total number of POIs, the number of the initialization population, the POI refinement category and the candidate POI data set, n initialization trips are randomly generated, where n is the number of the initialization population.

7. The personalized itinerary recommendation method according to claim 1, characterized in that: The step of screening the recommended itinerary to be screened according to the fitness value to obtain the recommended itinerary comprises: Performing non-dominated sorting on the recommended itineraries to be screened according to the fitness values, and selecting the first m recommended itineraries to be screened as itineraries to be recommended; Calculating the spatial distance between the recommended itinerary and a preset reference point; The to-be-recommended itinerary is associated with the nearest reference point according to the spatial distance to obtain the number of associations of each reference point, and a recommended itinerary is obtained by screening the recommended itineraries based on the number of associations.

8. A trip recommendation system, characterized in that: include: An acquisition module is used to acquire user travel big data of user travel, travel data of current users to travel, several POIs and POI data corresponding to the POIs; A classification module is used to classify several POIs and obtain POI detailed categories; An extraction module, used to analyze the user travel big data and extract the objects of interest that the user pays attention to when traveling; A prediction module, for predicting the number of overall POIs and entertainment POIs that the user wants to include in the itinerary, the probability of the user selecting a certain POI refinement category, the probability of the user paying attention to the target of interest, and the probability of the user preferring the fine-grained attributes of the target of interest under a certain POI refinement category, based on the travel data of the current user to be traveling, the Ci prediction model, the Ti prediction model, the Wi prediction model, and the Di prediction model; A first screening module, configured to screen POIs according to the preference probability and the focus target to obtain a candidate POI data set; Design module, used to design the initialization itinerary; A crossover mutation module is used to optimize the initialization trip through crossover mutation of a genetic algorithm according to the candidate POI data set, the total number of POIs, the number of entertainment POIs, the probability of the user selecting a certain POI refinement category, and the probability of the user paying attention to the attention target, to obtain several trips to be selected, and calculate the fitness value of the initialization trip and the fitness value of the trip to be selected, wherein the fitness value is calculated through the attention target; The initialized itinerary and the itinerary to be selected constitute a recommended itinerary to be screened; A second screening module, configured to screen the recommended itineraries to be screened according to the fitness value to obtain the recommended itineraries; The recommendation module is used to obtain the best recommended itinerary for the user's travel based on the recommended itinerary and the user preference probability.

9. A terminal device, comprising a memory and a processor, characterized in that: The memory stores a computer program that can be run on the processor, and when the processor loads and executes the computer program, the recommended method according to any one of claims 1 to 7 is adopted.

10. A computer-readable storage medium having a computer program stored therein, characterized in that: When the computer program is loaded and executed by a processor, the recommended method according to any one of claims 1 to 7 is adopted.

Images