Multi-objective tourism recommendation method and system based on neighborhood selection

Abstract

The present invention discloses a multi-objective tourism recommendation method based on neighborhood selection, which includes the following steps: S1: Dataset construction: Collect the historical access sequences of users and perform data screening on the dataset; S2: Construction of the user neighbor pool: Divide users into new users and old users, calculate the neighbor lists of different target users, and finally select the top m users with the highest similarity as neighbors and put them into the neighbor pool; S3: Establishment of a multi-objective recommendation model: Convert the two metrics of recommendation accuracy and novelty that users are concerned about into a multi-objective optimization model with two objectives; S4: Optimize the neighbor list using a multi-objective evolutionary algorithm: Use an improved multi-objective evolutionary algorithm to optimize the neighbor list to obtain the optimal neighbor list; S5: Generate a scenic spot recommendation list for the target user according to the optimal neighbor set obtained by optimization. Also disclosed is a multi-objective tourism recommendation system based on neighborhood selection. The present invention can obtain multiple non-dominated solutions that meet the requirements.

Description

Technical Field

[0001] The present invention relates to the field of optimization computing, and particularly to a multi-objective tourism recommendation method and system based on neighborhood selection. Background Art

[0002] With the improvement of residents' living standards, the requirements of residents for spiritual culture and a better life are becoming increasingly strong. Tourism has become an important part of people's leisure and entertainment, and the tourism industry has developed rapidly. Facing the huge and complex tourism data, people are easily overwhelmed by a large amount of data, and it is difficult for users to quickly find the tourist attractions they need. Therefore, in order to meet the increasingly urgent needs of people for efficient and intelligent tourism information, effectively solve the tourism recommendation service, and recommend more favorite attractions to users, an efficient tourism recommendation system is very necessary.

[0003] Recommendation systems focusing on accuracy tend to recommend popular attractions, while attractions with fewer visits are also important in practice. Therefore, we should not only actively explore users' interests but also pay attention to the novelty of recommendations. In addition, since the user-based collaborative filtering algorithm is currently the most popular recommendation method, which generates recommendations based on the neighborhood of users, selecting the correct neighbors has a great impact on the quality of recommendations.

[0004] Therefore, there is an urgent need to provide a new multi-objective tourism recommendation method and system based on neighborhood selection to solve the above problems. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a multi-objective tourism recommendation method and system based on neighborhood selection, which can optimize the accuracy and novelty of recommendations by using the multi-objective evolution characteristics of the evolutionary algorithm, and at the same time find the best neighbor list of the user-based collaborative filtering algorithm to obtain multiple non-dominated solutions that meet the requirements.

[0006] To solve the above technical problem, a technical solution adopted by the present invention is: to provide a multi-objective tourism recommendation method based on neighborhood selection, including the following steps:

[0007] S1: Dataset construction: Collect the historical access sequences of users, sort them according to the IDs and access times of user access, and screen the collected dataset.

[0008] S2: Construction of the user neighbor pool: Divide users into new users and old users according to the number of tourist attractions visited by users. For target users in different states, different calculation methods are used to calculate the neighbor list of the target user, and finally, the top m users with the highest similarity are selected as neighbors and put into the neighbor pool.

[0009] S3: Establishment of multi-objective recommendation model: Transform the performance of the scenic spot recommendation list concerned by users into a multi-objective optimization problem containing two optimization objectives through modeling, that is, transform the two indicators of recommendation accuracy and novelty into a multi-objective optimization model with two objectives, and calculate the objective function;

[0010] S4: Optimize the neighbor list using a multi-objective evolutionary algorithm: Use an improved multi-objective evolutionary algorithm to optimize the neighbor list to obtain the optimal neighbor list;

[0011] S5: Generation of user recommendation list: Obtain the non-dominated solution set of the user's neighbors according to the multi-objective evolutionary algorithm, select the optimal neighbor solution of the user from the non-dominated solution set, and calculate the scenic spot recommendation list of the user according to the user-based collaborative filtering algorithm.

[0012] In a preferred embodiment of the present invention, in step S1, data screening of the collected data set includes filtering out locations visited by fewer than at most 4 different users from the HSC data set and filtering out locations visited by fewer than at most 10 different users from the NYC data set.

[0013] In a preferred embodiment of the present invention, in step S2, the specific steps for calculating the neighbor list of the target user include:

[0014] (1) When the target user is an old user, the similarity between the target user and other users is calculated through the binary form of cosine similarity:

[0015]

[0016] where, I 11 represents the number of scenic spots jointly rated by user x and user u, I 10 represents the number of ratings given by user x but not given by user u, I 01 represents the number of ratings not given by user x but given by user u;

[0017] (2) When the target user is a new user, the similarity of scenic spots and OUsim similarity are jointly used to select the neighbors of the target user: First, generate a Top-20 scenic spot similarity list according to the scenic spot similarity, and all users who have rated at least one scenic spot in the scenic spot similarity list will be selected and added to the TopUser1 list; then calculate the OUsim between the target user and other users to generate the TopUser2 list; finally, generate an intersection user list AUL using these two lists, where the scenic spot similarity is:

[0018]

[0019] Among them, |N(i)| is the number of users who like scenic spot i. In the dataset, a true rating of a scenic spot exceeding 4 is defined as liking, and |N(i)∩N(j)| is the number of users who like both scenic spot i and scenic spot j;

[0020] AUL = |TopUser1∩TopUser2|

[0021] In a preferred embodiment of the present invention, in step S3, the steps of calculating the objective function include:

[0022] (1) The function for measuring the recommendation accuracy uses the item ratings predicted by traditional recommendation techniques. The objective function for measuring accuracy is:

[0023]

[0024] Among them, r u,i is the rating of user u for scenic spot i, and k represents the length of the scenic spot recommendation list. The larger the value of this objective function, the more popular the items in the scenic spot recommendation list;

[0025] (2) Novelty indicates the ability of the recommendation system to recommend scenic spots that are less noticed by users. Novelty is defined as:

[0026]

[0027] Among them, u i and σ i are the average value and variance of all users' scores for scenic spot i respectively. Then, when the scenic spot is more popular, the value is lower;

[0028] The finally constructed multi-objective function is:

[0029]

[0030] In a preferred embodiment of the present invention, the specific steps of step S4 include:

[0031] S4.1: Initialization of the population: Each individual in the population represents a neighbor list of length n. When initializing an individual, randomly select n neighbors from the neighbor pool of the target user for filling, and perform multiple individual initializations to complete the initialization of the population;

[0032] S4.2: Calculation of fitness: Select the population according to the fitness of the population. The fitness of the population is obtained from the two optimization objectives transformed by S3. The sum of the two objective functions indicates the size of the fitness. The population with a large fitness value enters the next generation;

[0033] S4.3: Crossover and mutation of the population: Use single-point uniform crossover and single-point mutation to generate new individuals respectively, and continuously update the population;

[0034] S4.4: If the current iteration count reaches the set maximum iteration count, end the optimization; otherwise, go to step S4.2 to continue the optimization.

[0035] Furthermore, in step S4.3, the specific steps of single-point uniform crossover include:

[0036] First, randomly select two parent individuals P1 and P2;

[0037] Then, randomly select a break point i on individuals P1 and P2, and exchange items between the end point i and end point j of individuals P1 and P2 to obtain offspring individuals O1 and O2. If duplicate items appear in the offspring individuals, use randomly selected candidate items to replace the duplicate items in the unchanged part of the individuals.

[0038] Furthermore, in step S4.3, the specific method of single-point mutation includes:

[0039] Determine the mutation point by random selection, and then use a candidate item that does not appear in the parent individual to replace the item at this mutation point in the offspring individual.

[0040] To solve the above technical problems, another technical solution adopted by the present invention is: to provide a multi-objective tourism recommendation system based on neighborhood selection, including:

[0041] A data construction module, which is used to collect the historical access sequences of users, sort them according to the user IDs and access times, and perform data screening on the collected data set;

[0042] A construction module for the user neighbor pool, which is used to divide users into new users and old users according to the number of scenic spots visited by users, calculate the similarity between the target user and other users using different similarity calculation methods for target users in different states, and finally select the top m users with the highest similarity as neighbors and put them into the neighbor pool;

[0043] A multi-objective recommendation model establishment module, which is used to construct a multi-objective optimization model for the two objectives of recommendation accuracy and novelty that users are concerned about, and calculate the objective function;

[0044] A multi-objective evolutionary algorithm optimization neighbor list module, which is used to optimize the neighbor list using an improved multi-objective evolutionary algorithm to obtain the optimal neighbor list;

[0045] A user recommendation list generation module, which is used to select the optimal neighbor solution of the user from the non-dominated solution set of the user neighbors that can be obtained by the multi-objective evolutionary algorithm, and calculate the user's recommendation list according to the user-based collaborative filtering algorithm.

[0046] In a preferred embodiment of the present invention, the steps of the multi-objective recommendation model establishing module for calculating the objective function include:

[0047] (1) The function for measuring the recommendation accuracy uses the item ratings predicted by traditional recommendation techniques. The objective function for measuring accuracy is:

[0048]

[0049] where r u,i is the rating given by user u to scenic spot i, and k represents the length of the recommendation list. The larger the value of this objective function, the more popular the items in the list are;

[0050] (2) Novelty indicates the ability of the recommendation system to recommend scenic spots that are less noticed by users. The definition of novelty is:

[0051]

[0052] where u i and σ i are the average value and variance of the scores given by all users to scenic spot i respectively. Then, when the scenic spot is more popular, its value is lower;

[0053] The finally constructed multi-objective function is:

[0054]

[0055] In a preferred embodiment of the present invention, the multi-objective evolutionary algorithm for optimizing the neighbor list module is used to optimize the neighbor list using an improved multi-objective evolutionary algorithm to obtain the optimal neighbor list. The specific steps include:

[0056] S4.1: Initialization of the population: Each individual in the population represents a neighbor list of length n. When initializing an individual, randomly select n neighbors from the neighbor pool of the target user for filling, and perform multiple individual initializations to complete the initialization of the population;

[0057] S4.2: Calculation of fitness: Select the population according to the fitness of the population. The fitness of the population is obtained from the two optimization objectives transformed by S3. The sum of the two objective functions indicates the size of the fitness. The population with a large fitness value enters the next generation;

[0058] S4.3: Crossover and mutation of the population: Use single-point uniform crossover and single-point mutation to generate new individuals respectively, and continuously update the population;

[0059] S4.4: If the current iteration number reaches the set maximum iteration number, end the optimization, otherwise go to step S4.2 to continue the optimization.

[0060] The beneficial effects of the present invention are as follows:

[0061] (1) To solve the conflict relationship between the two objectives of accuracy and novelty in the travel recommendation system, the present invention adopts a multi-objective optimization algorithm to handle the optimization problem among multiple conflicting objectives, improving the novelty of the recommendation on the basis of ensuring the recommendation accuracy, and greatly enhancing the user experience.

[0062] (2) The present invention uses a multi-objective evolutionary algorithm to optimize the user's neighbor list, which can help users find the optimal neighbor list, thereby improving the quality of the recommendation. In addition, compared with the traditional method that can only obtain one solution, this method can obtain multiple non-dominated solutions, contributing to the development of subsequent recommendation algorithms. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 is a schematic diagram of the overall framework process of the multi-objective travel recommendation method based on neighborhood selection of the present invention;

[0064] Figure 2 is a schematic diagram of the crossover operator in the multi-objective evolutionary algorithm;

[0065] Figure 3 is a schematic diagram of the mutation operator in the multi-objective evolutionary algorithm;

[0066] Figure 4 is a structural block diagram of the multi-objective travel recommendation system based on neighborhood selection. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0067] The following elaborates on the preferred embodiments of the present invention in detail with reference to the accompanying drawings, so that the advantages and features of the present invention can be more easily understood by those skilled in the art, thereby making a clearer and more definite definition of the protection scope of the present invention.

[0068] Please refer to Figure 1 , the embodiments of the present invention include:

[0069] S1: Dataset construction: Collect the historical access sequences of users from Foursquare and travel websites, sort them according to the user ID and access time, and screen the collected dataset.

[0070] Further, screening the collected data includes filtering out locations with less than at most 4 different user visits from the HSC dataset and filtering out locations with less than at most 10 different user visits from the NYC dataset.

[0071] S2: Steps for constructing the user neighbor pool: Users are divided into new users and old users according to the number of scenic spots they visit. Different calculation methods are used for target users in different states to calculate the neighbor list of the target user. Finally, the top 30 users with the highest similarity are selected as neighbors and put into the neighbor pool.

[0072] Furthermore, the specific steps for calculating the neighbor list of the target user are as follows:

[0073] (1) When the target user is an old user, the similarity between the target user and other users is calculated through the binary form of cosine similarity:

[0074]

[0075] where, I 11 is the number of scenic spots jointly rated by user x and user u, I 10 represents the number of scenic spots rated by user x but not rated by user u, I 01 represents the number of scenic spots not rated by user x but rated by user u.

[0076] (2) When the target user is a new user, since the number of scenic spots rated by new users is small, similar users may not be determined because there are not enough common ratings between users. Therefore, the similarity based on scenic spots and OUsim similarity are jointly used to select the neighbors of the target user.

[0077] First, a Top-20 scenic spot similarity list is generated based on scenic spot similarity. All users who have rated at least one scenic spot in the scenic spot similarity list will be selected and added to the TopUser1 list; then, the OUsim between the target user and other users is calculated to generate the TopUser2 list; finally, a cross-user list (AUL) is generated using these two lists, where the scenic spot similarity is:

[0078]

[0079] where, |N(i)| is the number of users who like scenic spot i. A scenic spot with a true rating of more than 4 in the dataset is defined as liked. |N(i)∩N(j)| is the number of users who like both scenic spot i and scenic spot j.

[0080] AUL = |TopUser1 ∩ TopUser2|

[0081] S3: Constructing a multi-objective optimization model: The performance of the scenic spot recommendation list that the user is concerned about is transformed into a multi-objective optimization problem with two optimization objectives through modeling, that is, the two indicators of recommendation accuracy and novelty are transformed into a multi-objective optimization model with two objectives, and the objective function is calculated.

[0082] The steps for further calculating the objective function include: To give a function for measuring the recommendation accuracy, the item ratings predicted by traditional recommendation techniques are used. To recommend items that users like, the items in the scenic spot recommendation list should have higher ratings. Then the objective function for measuring accuracy is:

[0083]

[0084] where r u,i is the rating given by user u to scenic spot i, k represents the length of the scenic spot recommendation list, and the larger the value of this objective function, the more popular the items in the scenic spot recommendation list are.

[0085] Novelty indicates the ability of the recommendation system to recommend scenic spots that are less noticed by users. The more niche items there are in the recommendation list, the higher the novelty obtained by the recommendation algorithm. Novelty can be defined as:

[0086]

[0087] where u i and σ i are the mean and variance of all users' ratings for scenic spot i respectively. Then when the scenic spot is more popular, the value is lower.

[0088] The finally constructed multi-objective function is:

[0089]

[0090] S4: Optimize the neighbor list using a multi-objective evolutionary algorithm: Use an improved multi-objective evolutionary algorithm to optimize the neighbor list to obtain the optimal neighbor list. The specific steps include:

[0091] S4.1: Initialization of the population: Each individual in the population represents a neighbor list of length 10. When initializing an individual, randomly select 10 neighbors from the neighbor pool of the target user for filling. Perform multiple individual initializations to complete the initialization of the population;

[0092] S4.2: Calculation of fitness: Select the population according to the fitness of the population. The fitness of the population is obtained from the two optimization objectives transformed in S3. The sum of the two objective functions indicates the size of the fitness, and the population with a large fitness value enters the next generation;

[0093] S4.3: Crossover and mutation of the population: The multi-objective evolutionary algorithm often uses simulated binary crossover for real number coding, but this crossover method is not suitable for the individual coding method of this method. Therefore, single-point uniform crossover is implemented, as Figure 2As shown below. The specific process of single-point uniform crossover is as follows: First, to improve the diversity of the population and explore a larger search space, two parent individuals P1 and P2 are randomly selected; then, a breakpoint i is randomly selected on individuals P1 and P2, and the items between endpoint i and endpoint j of individuals P1 and P2 are exchanged with each other to obtain offspring individuals O1 and O2, that is Figure 2 The 1 in it transitions to 2. Through the above steps, offspring individuals O1 and O2 can be obtained. Using single-point uniform crossover may cause duplicate items to appear in the offspring individuals. For example, Figure 2 the offspring individual O1 in it has a duplicate item 24, and O2 has a duplicate item 30. To solve this problem, the process of single-point crossover is extended, and randomly selected candidate items are used to replace the duplicate items in the unchanged part of the individual. For example, the duplicate item 24 in the unchanged part of individual O1 is randomly replaced with 23, and the duplicate item 30 in the unchanged part of individual O2 is randomly replaced with 29.

[0094] At the same time, single-point mutation is used to generate new individuals. As shown in Figure 3 below, the mutation point is determined by random selection. As shown in Figure 3 the dashed box in it, the random probability of each item in the individual is 1 / 10; then an item that does not appear in the parent individual is used to replace the item at the mutation point. For example, Figure 3 the 2 in individual P mutates to 11 in individual O. The single-point mutation operation is simple and has a low computational complexity, which can effectively ensure the diversity of individuals.

[0095] S4.4: If the current iteration number reaches the set maximum iteration number, the optimization ends; otherwise, go to step S4.2 to continue the optimization.

[0096] S5: Generation of the recommendation list: According to the multi-objective evolutionary algorithm, the non-dominated solution set of the user's neighbors can be obtained. The user's optimal neighbor solution is selected from the non-dominated solution set, and the scenic spot recommendation list for the user is calculated based on the user-based collaborative filtering algorithm.

[0097] Refer to Figure 4 , in the example of the present invention, a multi-objective tourism recommendation system based on neighborhood selection is also provided, including:

[0098] A data construction module, which is used to collect the historical access sequences of users, sort them according to the ID and access time of the users' accesses, and perform data screening on the collected data set;

[0099] A construction module for the user neighbor pool, which is used to divide users into new users and old users according to the number of scenic spots visited by the users. Different similarity calculation methods are used for target users in different states to calculate the similarity between the target user and other users. Finally, the top m users with the highest similarity are selected as neighbors and put into the neighbor pool; in this example, m is taken as 30;

[0100] A multi-objective recommendation model building module, which is used to construct a multi-objective optimization model with the two objectives of recommendation accuracy and novelty that the user is concerned about, and calculate the objective function;

[0101] A multi-objective evolutionary algorithm optimization neighbor list module, which is used to optimize the neighbor list using an improved multi-objective evolutionary algorithm to obtain an optimal neighbor list;

[0102] A user recommendation list generation module, which is used to select the optimal neighbor solution of the user from the non-dominated solution set of the user's neighbors that can be obtained according to the multi-objective evolutionary algorithm, and calculate the user's recommendation list according to the user-based collaborative filtering algorithm.

[0103] Among them, the steps for the multi-objective recommendation model building module to calculate the objective function include:

[0104] (1) The function for measuring recommendation accuracy uses the item ratings predicted by traditional recommendation techniques. The objective function for measuring accuracy is:

[0105]

[0106] Among them, r u,i is the rating of scenic spot i by user u, and k represents the length of the recommendation list. The larger the value of this objective function, the more popular the items in the list;

[0107] (2) Novelty indicates the ability of the recommendation system to recommend scenic spots that are less noticed by users. The definition of novelty is:

[0108]

[0109] Among them, u i and σ i are the average and variance of all users' ratings of scenic spot i respectively. Then, when the scenic spot is more popular, its value is lower;

[0110] The finally constructed multi-objective function is:

[0111]

[0112] Among them, the multi-objective evolutionary algorithm optimization neighbor list module, which is used to optimize the neighbor list using an improved multi-objective evolutionary algorithm to obtain an optimal neighbor list, the specific steps include:

[0113] S4.1: Initialization of the population: Each individual in the population represents a neighbor list of length n. When initializing the individual, randomly select n neighbors from the neighbor pool of the target user for filling, and perform multiple individual initializations to complete the initialization of the population; In this example, n is taken as 10;

[0114] S4.2: Calculation of fitness: Select the population according to the fitness of the population. The fitness of the population is obtained from the two optimization objectives transformed in S3. The magnitude of the sum of the two objective functions indicates the fitness. The population with a large fitness value enters the next generation;

[0115] S4.3: Crossover and mutation of the population: New individuals are generated by single-point uniform crossover and single-point mutation respectively, and the population is continuously updated;

[0116] S4.4: If the current iteration number reaches the set maximum iteration number, the optimization ends. Otherwise, go to step S4.2 to continue the optimization.

[0117] The above are only the embodiments of the present invention, and do not limit the patent scope of the present invention accordingly. Any equivalent structure or equivalent process transformation made by using the content of the specification and drawings of the present invention, or directly or indirectly applied in other related technical fields, shall be equally included in the patent protection scope of the present invention.

Claims

1. A multi-objective tourism recommendation method based on neighborhood selection, characterized in that Including the following steps: S1: Dataset construction: Collect the historical access sequences of users, sort them according to the user IDs and access times, and perform data screening on the collected dataset; S2: Construction of user neighbor pool: Divide users into new users and old users according to the number of scenic spots visited by users. For target users in different states, use different calculation methods to calculate the neighbor lists of target users, and finally select the top m users with the highest similarity as neighbors and put them into the neighbor pool; S3: Establishment of multi-objective recommendation model: Transform the performance of the scenic spot recommendation list concerned by users into a multi-objective optimization problem containing two optimization objectives through modeling, that is, transform the two indicators of recommendation accuracy and novelty into a multi-objective optimization model with two objectives, and calculate the objective function; The steps of calculating the objective function include: (1) The function for measuring recommendation accuracy uses the item ratings predicted by traditional recommendation techniques, and the objective function for measuring accuracy is: where r u,i is the rating given by user u to attraction i, and k represents the length of the attraction recommendation list. The larger the value of this objective function, the more popular the items in the attraction recommendation list are; (2) Novelty indicates the ability of the recommendation system to recommend scenic spots that are less noticed by users, and novelty is defined as: where u i and σ i are the mean and variance of the scores given by all users to attraction i, respectively. Then, the more popular the attraction is, the lower its value is; The finally constructed multi-objective function is: S4: Optimize the neighbor list using a multi-objective evolutionary algorithm: Use an improved multi-objective evolutionary algorithm to optimize the neighbor list to obtain the optimal neighbor list; S5: Generation of user recommendation list: Obtain the non-dominated solution set of user neighbors according to the multi-objective evolutionary algorithm, select the optimal neighbor solution of the user from the non-dominated solution set, and calculate the scenic spot recommendation list of the user according to the user-based collaborative filtering algorithm.

2. The multi-objective tourism recommendation method based on neighborhood selection according to claim 1, wherein In step S1, the data screening of the collected dataset includes filtering out locations visited by fewer than at most 4 different users from the HSC dataset and filtering out locations visited by fewer than at most 10 different users from the NYC dataset.

3. The multi-objective tourism recommendation method based on neighborhood selection according to claim 1, characterized in that In step S2, the specific steps of calculating the neighbor list of the target user include: (1) When the target user is an old user, the similarity between the target user and other users is calculated through the binary form of cosine similarity: Among them, I 11 represents the number of scenic spots jointly rated by user x and user u, I 10 represents the number of ratings given by user x that user u did not give, I 01 represents the number of ratings not given by user x that user u gave; (2) When the target user is a new user, the similarity of scenic spots and OUsim similarity are used together to select the neighbors of the target user: First, generate a Top-20 scenic spot similarity list according to scenic spot similarity, and all users who have rated at least one scenic spot in the scenic spot similarity list will be selected and added to the TopUser1 list; Then calculate the OUsim between the target user and other users to generate the TopUser2 list; Finally, use these two lists to generate a cross-user list AUL, where the scenic spot similarity is: where |N(i)| is the number of users who like scenic spot i, and a true rating of more than 4 for a scenic spot in the dataset is defined as like, and |N(i)∩N(j)| is the number of users who like both scenic spot i and scenic spot j; AUL = |TopUser1∩TopUser2|.

4. The multi-objective tourism recommendation method based on neighborhood selection according to claim 1, characterized in that The specific steps of step S4 include: S4.1: Initialization of the population: Each individual in the population represents a neighbor list of length n. When initializing an individual, randomly select n neighbors from the neighbor pool of the target user for filling. Perform multiple individual initializations to complete the initialization of the population; S4.2: Calculation of fitness: Select the population according to the fitness of the population. The fitness of the population is obtained from the two optimization objectives transformed by S3. The sum of the two objective functions indicates the fitness. The population with a larger fitness value enters the next generation; S4.3: Crossover and mutation of the population: Use single-point uniform crossover and single-point mutation to generate new individuals respectively, and continuously update the population; S4.4: If the current iteration number reaches the set maximum iteration number, end the optimization, otherwise go to step S4.2 to continue the optimization.

5. The multi-objective tourism recommendation method based on neighborhood selection according to claim 4, characterized in that In the step S4.3, the specific steps of single-point uniform crossover include: First, randomly select two parent individuals P1 and P2; Then, randomly select a break point i on individuals P1 and P2, and exchange items between the endpoints i to j of individuals P1 and P2 to obtain offspring individuals O1 and O2. If there are duplicate items in the offspring individuals, use randomly selected candidate items to replace the duplicate items in the unchanged part of the individuals.

6. The multi-objective tourism recommendation method based on neighborhood selection according to claim 4, wherein In the step S4.3, the specific method of single-point mutation includes: Determine the mutation point by random selection, and then use a candidate item that does not appear in the parent individual to replace the item at this mutation point in the offspring individual.

7. A multi-objective tourism recommendation system based on neighborhood selection, characterized in that, Include: Data construction module, used to collect the historical access sequences of users, sort them according to the ID and access time of user access, and perform data screening on the collected data set; Module for constructing the user neighbor pool, used to divide users into new users and old users according to the number of scenic spots visited by users, calculate the similarity between the target user and other users using different similarity calculation methods for target users in different states, and finally select the top m users with the highest similarity as neighbors and put them into the neighbor pool; Module for establishing a multi-objective recommendation model, used to construct a multi-objective optimization model with the two objectives of recommendation accuracy and novelty that the user is concerned about, and calculate the objective function; Module for optimizing the neighbor list using a multi-objective evolutionary algorithm, used to optimize the neighbor list using an improved multi-objective evolutionary algorithm to obtain the optimal neighbor list; Module for generating the user recommendation list, used to select the optimal neighbor solution of the user from the non-dominated solution set of the user neighbors that can be obtained by the multi-objective evolutionary algorithm, and calculate the user's recommendation list according to the user-based collaborative filtering algorithm.

8. The multi-objective tourism recommendation system based on neighborhood selection according to claim 7, characterized in that The steps for the module for establishing a multi-objective recommendation model to calculate the objective function include: (1) The function for measuring recommendation accuracy uses the item ratings predicted by traditional recommendation techniques. The objective function for measuring accuracy is: where r u,i is the rating given by user u to attraction i, and k represents the length of the recommendation list. The larger the value of this objective function, the more popular the items in the list are; (2) Novelty indicates the ability of the recommendation system to recommend scenic spots that are less noticed by users. Novelty is defined as: where u i and σ i are the mean and variance of all users' ratings for attraction i, respectively. Then, the more popular the attraction, the lower its value. The finally constructed multi-objective function is: 。 9. The multi-objective tourism recommendation system based on neighborhood selection according to claim 7, characterized in that The module for optimizing the neighbor list using a multi-objective evolutionary algorithm, used to optimize the neighbor list using an improved multi-objective evolutionary algorithm to obtain the optimal neighbor list, the specific steps include: S4.1: Initialization of the population: Each individual in the population represents a neighbor list of length n. When initializing an individual, randomly select n neighbors from the neighbor pool of the target user for filling. Perform multiple individual initializations to complete the initialization of the population; S4.2: Calculation of fitness: Select the population according to the fitness of the population. The fitness of the population is obtained from the two optimization objectives transformed by S3. The sum of the two objective functions indicates the fitness. The population with a larger fitness value enters the next generation; S4.3: Crossover and mutation of the population: Use single-point uniform crossover and single-point mutation to generate new individuals respectively, and continuously update the population; S4.4: If the current iteration number reaches the set maximum iteration number, end the optimization, otherwise go to step S4.2 to continue the optimization.

Images