Multi-model fusion tourism recommendation method and system based on multi-objective optimization selection strategy

Abstract

The invention discloses a multi-model fusion tourism recommendation method based on a multi-objective optimization selection strategy, comprising the following steps: S1: selecting two or more single models in a tourism recommendation problem, and obtaining the recommendation results of each single recommendation model to obtain a total set to be optimized; S2: modeling the multi-model fusion recommendation problem as a multi-objective optimization problem; S3: constructing the total set to be optimized into multiple sub-sets to be optimized of a tourism recommendation scheme; S4: using a multi-objective optimization method to calculate the order of individuals in all sub-sets to be optimized to obtain an optimized sub-set; S5: merging all optimized sub-sets, and selecting the top k scenic spots in the optimized total set as the final output result of the model according to demand, that is, obtaining the final tourism recommendation result. A multi-model fusion tourism recommendation system based on a multi-objective optimization selection strategy is also disclosed. The invention integrates two or more sub-recommendation models to improve the accuracy of tourism recommendations.

Description

Technical Field

[0001] The present invention relates to the technical field of tourism recommendation, and in particular to a multi-model fusion tourism recommendation method and system based on a multi-objective optimization selection strategy. Background Art

[0002] In recent years, with the steady growth of the economy, the tourism industry has become more and more prosperous. At the same time, with the advancement of Internet technology, smart tourism has begun to receive more and more attention. During daily travel, people are accustomed to using the Internet to query relevant travel information and make their own travel plans. However, the explosive growth of data makes it very difficult for people to query valuable information in massive data, and a lot of time is spent on searching for travel information, which makes people's travel experience sharply reduced. Therefore, it is very important to develop an intelligent and accurate travel recommendation method.

[0003] In the existing technical background, content-based recommendation, collaborative filtering and RNN algorithms have dominated the mainstream. However, these algorithms have their own limitations. For example, content-based methods recommend based on the content features of items, collaborative filtering recommends based on the similarity of items or users, and RNN recommends based on the user's access time series. In addition, the existing soft voting and hard voting fusion methods must require the fusion of more than three sub-models.

[0004] Therefore, there is an urgent need to provide a new multi-model fusion tourism recommendation method and system based on multi-objective optimization selection strategy 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-model fusion tourism recommendation method and system based on a multi-objective optimization selection strategy, which can overcome the shortcomings of a single model, absorb the advantages of each single model, improve the accuracy of the recommendation results, and increase the flexibility of model fusion.

[0006] In order to solve the above technical problems, a technical solution adopted by the present invention is to provide a multi-model fusion tourism recommendation method based on a multi-objective optimization selection strategy, comprising the following steps:

[0007] S1: Select two or more single models in the tourism recommendation problem, obtain the recommendation results of each single recommendation model, and merge all the results to obtain the total set to be optimized;

[0008] S2: Model the multi-model fusion recommendation problem as a multi-objective optimization problem;

[0009] S3: constructing the total set to be optimized into multiple sub-sets to be optimized of travel recommendation solutions;

[0010] S4: Use the multi-objective optimization method to calculate the ranking of individuals in all subsets to be optimized and obtain the optimized subsets;

[0011] S5: All optimized sub-sets are merged in the reverse order of the sorting of the sub-sets to be optimized to obtain the optimized total set. The first k attractions are selected from the optimized total set as the final output result of the model according to the needs, that is, the final tourism recommendation result is obtained.

[0012] In a preferred embodiment of the present invention, the specific steps of step S2 include:

[0013] S2.1: Based on the historical access records S of user u u ={x 1 , x 2 ,......,x t}, x i represents the scenic spot visited by user i for the first time, select z (z ≥ 2) single recommendation models to predict the scenic spot visited by user u for the t+1th time, and obtain a set of z recommendation sequences with a length of k, denoted as L = {L1, L2, ..., L z}, L i represents the recommendation result of the i-th sub-recommendation model, where represents the jth scenic spot recommended by the i-th sub-recommendation model to user u, i∈{1,2,…,z},j∈{1,2,…,k}, and the results of each sub-recommendation model in L are integrated to construct a recommendation set X containing z*k scenic spot objects. u ;

[0014] S2.2: Will be from X u The problem of selecting k scenic spots is converted into a multi-objective optimization selection problem Maximize F(x) = (f1(x), f2(x), f3(x)), where f1(x), f2(x), and f3(x) are three different objective functions:

[0015] f1(x) represents the user’s interest in attraction x. i∈{1,2,……,t},x i ∈S u , where sim(x,x i ) represents the recommended scenic spot x and the i-th scenic spot x visited by user u i The similarity of U x represents the set of users who visit attraction x, Indicates visiting attractions x i The user set of |U x | represents the total number of people visiting attraction x, Indicates visiting attractions xi Total number of people;

[0016] f2(x) represents the popularity of attraction x, v x represents the score of scenic spot x, f2(x) is normalized by min-max;

[0017] f3(x) represents the time between scenic spot x and the last time the user visited scenic spot x t The inverse of the distance between D(x,x t ) represents the relationship between scenic spot x and scenic spot x t The distance between them is normalized by min-max.

[0018] In a preferred embodiment of the present invention, the specific steps of step S3 include:

[0019] S3.1: Statistics of the total set X to be optimized u The number of times each attraction x is repeated;

[0020] S3.2: Divide all scenic spots with a repetition count of 1 into subsets to be optimized Repeat this step until all scenic spots with a repetition number of z are divided into subsets to be optimized. A total of z subsets to be optimized are obtained, and the z subsets to be optimized are constructed into a set

[0021] Furthermore, the specific steps of step S4 include:

[0022] S4.1: For any subset in step S3.2 i∈{1, 2, ..., z}, express A scenic spot in The number of scenic spots in the map is sorted by non-dominated sorting method. The calculation formula is as follows: for any x1 and x2, if the relationship in formula (5) is satisfied, it means that x2 dominates x1, f i 、f j All of them are objective functions, F represents the objective function set, and n Pareto fronts are obtained through non-dominated sorting: Front = {Front1, Front2, ..., Front n}, Front i represents the i-th Pareto frontier, and n represents the number of Pareto frontiers;

[0023]

[0024] S4.2: For any Pareto front in S4.1 i ∈Front,i∈{1,2,...,n}, l≤j, for Front i For each individual x in the matrix, let the crowding distance of the border individual be infinite, and calculate the crowding distance of other individuals using formula (6), where x d The initial value of is 0, x+1 and x-1 represent the two individuals before and after x, and f i represents the objective function, |F| represents the total number of objective functions, and finally Frout is sorted from large to small according to the crowding distance. i Sort each individual in ;

[0025]

[0026] S4.3: Combine steps S4.1 and S4.2 to calculate Sorting all individuals in the set to obtain the optimized subset.

[0027] Furthermore, the specific steps of step S5 include:

[0028] S5.1: Calculate according to step S4 The internal individual sorting of all subsets to be optimized in Each optimized subset of Sort in order, and finally get The sorted set of all individuals after optimization, that is, the total set after optimization X′ u ;

[0029] S5.2: According to the demand, at X′ u The first k scenic spots are selected as the final output results of the model.

[0030] In order to solve the above technical problems, another technical solution adopted by the present invention is to provide a multi-model fusion tourism recommendation system based on a multi-objective optimization selection strategy, comprising:

[0031] The multi-model recommendation result calculation module is used to select two or more single models in the tourism recommendation problem, obtain the recommendation results of each single recommendation model, and merge all the results to obtain the total set to be optimized;

[0032] Multi-objective optimization model building module, used to model the multi-model fusion recommendation problem as a multi-objective optimization problem;

[0033] A module for constructing sub-sets to be optimized, used for constructing the total set to be optimized into multiple sub-sets to be optimized of travel recommendation solutions;

[0034] A multi-objective optimization calculation module is used to calculate the ranking of individuals in all subsets to be optimized using a multi-objective optimization method to obtain the optimized subsets;

[0035] The final tourism recommendation result output module is used to merge all optimized sub-sets in the reverse order of the sorting of the sub-sets to be optimized to obtain the optimized total set, and select the first k attractions in the optimized total set as the final output result of the model according to the needs, that is, to obtain the final tourism recommendation result.

[0036] In a preferred embodiment of the present invention, the specific steps of building a model of the multi-objective optimization model module include:

[0037] First, based on the historical access records S of user u u ={x 1 , x 2 ,......,x t}, x i represents the scenic spot visited by user i for the first time, select z (z ≥ 2) single recommendation models to predict the scenic spot visited by user u for the t+1th time, and obtain a set of z recommendation sequences with a length of k, denoted as L = {L1, L2, ..., L z}, L i represents the recommendation result of the i-th sub-recommendation model, where represents the jth scenic spot recommended by the i-th sub-recommendation model to user u, i∈{1,2,…,z},j∈{1,2,…,k}, and the results of each sub-recommendation model in L are integrated to construct a recommendation set X containing z*k scenic spot objects. u ;

[0038] Secondly, from X u The problem of selecting k scenic spots is converted into a multi-objective optimization selection problem Maximize F(x) = (f1(x), f2(x), f3(x)), where f1(x), f2(x), and f3(x) are three different objective functions:

[0039] f1(x) represents the user’s interest in attraction x. i∈{1, 2, ..., t}, x i ∈S u , where sim(x,x i ) represents the recommended scenic spot x and the i-th scenic spot x visited by user u i The similarity of U x represents the set of users who visit attraction x, Indicates visiting attractions x i The user set of |U x| represents the total number of people visiting attraction x, Indicates visiting attractions x i Total number of people;

[0040] f2(x) represents the popularity of attraction x, v x represents the score of scenic spot x, f2(x) is normalized by min-max;

[0041] f3(x) represents the time between scenic spot x and the last time the user visited scenic spot x t The inverse of the distance between D(x,x t ) represents the relationship between scenic spot x and scenic spot x t The distance between them is normalized by min-max.

[0042] In a preferred embodiment of the present invention, the specific steps of the to-be-optimized subset construction module constructing the to-be-optimized total set into multiple to-be-optimized subsets of travel recommendation solutions include:

[0043] First, count the total set X to be optimized u The number of times each attraction x is repeated;

[0044] Secondly, divide all scenic spots with a repetition number of 1 into subsets to be optimized Repeat this step until all scenic spots with a repetition number of z are divided into subsets to be optimized. A total of z subsets to be optimized are obtained, and the z subsets to be optimized are constructed into a set

[0045] In a preferred embodiment of the present invention, the specific calculation steps of the multi-objective optimization calculation module include:

[0046] First, for any subset i∈{1, 2, ..., z}, express A scenic spot in The number of scenic spots in the map is sorted by non-dominated sorting method. The calculation formula is as follows: for any x1 and x2, if the relationship in formula (5) is satisfied, it means that x2 dominates x1, f i 、f j All of them are objective functions, F represents the objective function set, and n Pareto fronts are obtained through non-dominated sorting: Front = {Front1, Front2, ..., Front n}, Front irepresents the i-th Pareto frontier, and n represents the number of Pareto frontiers;

[0047]

[0048] Secondly, for any of the above Pareto fronts Front i ∈Front,i∈{1,2,...,n}, l≤j, for Front i For each individual x in the matrix, let the crowding distance of the border individual be infinite, and calculate the crowding distance of other individuals using formula (6), where x d The initial value of is 0, x+1 and x-1 represent the two individuals before and after x, and f i represents the objective function, |F| represents the total number of objective functions, and finally Frout is sorted from large to small according to the crowding distance. i Sort each individual in ;

[0049]

[0050] Finally, combining the above two steps to calculate Sorting all individuals in the set to obtain the optimized subset.

[0051] In a preferred embodiment of the present invention, the final travel recommendation result output module calculates The internal individual sorting of all subsets to be optimized in Each optimized subset in Sort in order, and finally get The sorted set of all individuals after optimization, that is, the total set after optimization X′ u , according to the needs in X′ u The first k scenic spots are selected as the final output results of the model.

[0052] The beneficial effects of the present invention are:

[0053] (1) This invention proposes a multi-model fusion tourism recommendation method based on a multi-objective optimization selection strategy for the tourism recommendation problem, which can achieve higher recommendation accuracy;

[0054] (2) The present invention overcomes the limitations of existing single models and absorbs the advantages of each single model, which not only improves the accuracy of recommendation results, but also can adapt to more complex situations and has higher model generalization;

[0055] (3) The present invention can realize the fusion of two or more sub-models, making model fusion more flexible. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 It is a flow chart of a multi-model fusion tourism recommendation method based on a multi-objective optimization selection strategy of the present invention;

[0057] Figure 2 It is a structural block diagram of the multi-model fusion tourism recommendation system based on multi-objective optimization selection strategy. DETAILED DESCRIPTION

[0058] The preferred embodiments of the present invention are described in detail below in conjunction with 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.

[0059] See also Figure 1 , the embodiment of the present invention includes:

[0060] A multi-model fusion tourism recommendation method based on a multi-objective optimization selection strategy includes the following steps:

[0061] S1: Select two or more single models in the tourism recommendation problem, obtain the recommendation results of each single recommendation model, and merge all the results to obtain the total set to be optimized;

[0062] S1.1: For the travel recommendation database, let X = {x1, x2, ..., x n} represents the scenic spot data set, x i represents the i-th scenic spot, n is the total number of scenic spots in the database, i∈{1,2,...,n}; let U={u1,u2,...,u m} represents the user data set, u i represents the i-th user, m is the total number of users in the database, i∈{1, 2, ..., m}; let S u ={x1, x2, ..., x t} represents the sequence of scenic spots visited by user u, x i represents the i-th scenic spot visited by user u, t is the total number of scenic spots visited by user u, i∈{1,2,...,t}. Let M={M1,M2,...,M z} represents a set of z single recommendation models, M i represents the i-th sub-recommendation model, z is the total number of single recommendation models, i∈{1, 2, ..., z};

[0063] S1.2: The project-based collaborative filtering recommendation method, the user-based collaborative filtering recommendation method, and the bidirectional long short-term memory network (Bi-LSTM) recommendation method are selected as three basic single models. Users can also select or add other single models according to the actual situation they face;

[0064] S1.3: For all attractions in X, the similarity between the attractions is calculated according to the user access data of each attraction. The similarity between any two attractions x1 and x2 is calculated as shown in formula (1): represents all users who visited attraction x1, represents the total number of people visiting attraction x1, represents a user who visits x1 and x2 at the same time. According to formula (1), the similarity between any two points in X can be calculated. For any user u, according to his historical visit records S u ={x 1 , x 2 ,......,x t}, calculate the similarity between any scenic spot x and its historical visit records, and refer to formula (2) for the calculation method. Sort the historical similarity between all scenic spots and users, and select the top k scenic spots with the largest similarity as the final recommendation result L1 of the project-based collaborative filtering recommendation method, L1 = {x1, x2, ..., x k};

[0065]

[0066]

[0067] S1.4: For each user in U, according to the access record data between the users, the user similarity between any two users u1 and u2 is calculated according to formula (3). Indicates the historical scenic spot data visited by user u1. represents the total number of attractions visited by user u1, Indicates the attractions that users u1 and u2 have visited together. For any user u∈U, the similarity between u and all users is calculated by formula (3). The first q users are selected according to the user similarity from large to small, and the similarity between user u and these users is used as the weight to assign the historical attractions visited by q users. The weights of the same attractions are then accumulated to obtain a list to be recommended. Finally, the attractions in the list to be recommended are sorted from large to small according to the weight, and the first k attractions are selected as the output L2 of the user-based collaborative filtering recommendation method;

[0068]

[0069]

[0070] S1.5: For each user in U, input their historical visited scenic spots data into the BiLSTM neural network to obtain the third recommendation result L3.

[0071] S2: Model the multi-model fusion recommendation problem as a multi-objective optimization problem;

[0072] S2.1: Based on the historical access records S of user u u ={x 1 , x 2 ,......,x t}, x i represents the scenic spot visited by user i for the first time, select z (z ≥ 2) single recommendation models to predict the scenic spot visited by user u for the t+1th time, and obtain a set of z recommendation sequences with a length of k, denoted as L = {L1, L2, ..., L z}, L i represents the recommendation result of the i-th sub-recommendation model, where represents the jth scenic spot recommended by the i-th sub-recommendation model to user u, i∈{1,2,…,z},j∈{1,2,…,k}, and the results of each sub-recommendation model in L are integrated to construct a recommendation set X containing z*k scenic spot objects. u ;

[0073] S2.2: Will be from X u The problem of selecting k scenic spots is converted into a multi-objective optimization selection problem Maximize F(x) = (f1(x), f2(x), f3(x)), where f1(x), f2(x), and f3(x) are three different objective functions:

[0074] f1(x) represents the user’s interest in attraction x. i∈{1, 2, ..., t}, x i ∈S u , where sim(x,x i ) represents the recommended scenic spot x and the i-th scenic spot x visited by user u i The similarity of U x represents the set of users who visit attraction x, Indicates visiting attractions x i The user set of |U x | represents the total number of people visiting attraction x, Indicates visiting attractions x i Total number of people;

[0075] f2(x) represents the popularity of attraction x, v x represents the score of scenic spot x, f2(x) is normalized by min-max;

[0076] f3(x) represents the time between scenic spot x and the last time the user visited scenic spot x t The inverse of the distance between D(x,x t ) represents the relationship between scenic spot x and scenic spot x t The distance between them is normalized by min-max.

[0077] S3: constructing the total set to be optimized into multiple sub-sets to be optimized of travel recommendation solutions;

[0078] S3.1: Statistics of the total set X to be optimized u The number of times each attraction x is repeated;

[0079] S3.2: Divide all scenic spots with a repetition count of 1 into subsets to be optimized Repeat this step until all scenic spots with a repetition number of z are divided into subsets to be optimized. A total of z subsets to be optimized are obtained, and the z subsets to be optimized are constructed into a set

[0080] S4: Use multi-objective optimization methods to calculate the ranking of individuals in all subsets to be optimized;

[0081] S4.1: For any subset in step S3.2 i∈{1, 2, ..., z}, express A scenic spot in The number of scenic spots in the map is sorted by non-dominated sorting method. The calculation formula is as follows: for any x1 and x2, if the relationship in formula (5) is satisfied, it means that x2 dominates x1, f i 、f j All of them are objective functions, F represents the objective function set, and n Pareto fronts are obtained through non-dominated sorting: Front = {Front1, Front2, ..., Front n}, Front i represents the i-th Pareto frontier, and n represents the number of Pareto frontiers;

[0082]

[0083] S4.2: For any Pareto front in S4.1 i∈Front,i∈{1,2,...,n}, l≤j, for Front i For each individual x in the matrix, let the crowding distance of the border individual be infinite, and calculate the crowding distance of other individuals using formula (6), where x d The initial value of is 0, x+1 and x-1 represent the two individuals before and after x, and f i represents the objective function, |F| represents the total number of objective functions, and finally the Front i Sort each individual in ;

[0084]

[0085] S4.3: Combine steps S4.1 and S4.2 to calculate Sorting all individuals in the set to obtain the optimized subset.

[0086] S5: Merge all optimized sub-sets in the reverse order of the sorting of the sub-sets to be optimized to obtain the optimized total set. Select the first k attractions in the optimized total set as the final output result of the model according to the needs, that is, the final tourism recommendation result is obtained:

[0087] S5.1: Calculate according to step S4 The internal individual sorting of all subsets to be optimized in Each optimized subset of Sort in order, and finally get The sorted set of all individuals after optimization, that is, the total set after optimization X′ u ;

[0088] S5.2: According to the demand, at X′ u The first k scenic spots are selected as the final output results of the model.

[0089] See also Figure 2 The example of the present invention also provides a multi-model fusion tourism recommendation system based on a multi-objective optimization selection strategy, including:

[0090] The multi-model recommendation result calculation module is used to select two or more single models in the tourism recommendation problem, obtain the recommendation results of each single recommendation model, and merge all the results to obtain the total set to be optimized;

[0091] Multi-objective optimization model building module, used to model the multi-model fusion recommendation problem as a multi-objective optimization problem;

[0092] A module for constructing sub-sets to be optimized, used for constructing the total set to be optimized into multiple sub-sets to be optimized of travel recommendation solutions;

[0093] A multi-objective optimization calculation module is used to calculate the ranking of individuals in all subsets to be optimized using a multi-objective optimization method to obtain the optimized subsets;

[0094] The final tourism recommendation result output module is used to merge all optimized sub-sets in the reverse order of the sorting of the sub-sets to be optimized to obtain the optimized total set, and select the first k attractions in the optimized total set as the final output result of the model according to the needs, that is, to obtain the final tourism recommendation result.

[0095] Furthermore, the specific steps of establishing the multi-objective optimization model module modeling include:

[0096] First, based on the historical access records S of user u u ={x 1 , x 2 ,......,x t}, x i represents the scenic spot visited by user i for the first time, select z (z ≥ 2) single recommendation models to predict the scenic spot visited by user u for the t+1th time, and obtain a set of z recommendation sequences with a length of k, denoted as L = {L1, L2, ..., L z}, L i represents the recommendation result of the i-th sub-recommendation model, where represents the jth scenic spot recommended by the i-th sub-recommendation model to user u, i∈{1,2,…,z},j∈{1,2,…,k}, and the results of each sub-recommendation model in L are integrated to construct a recommendation set X containing z*k scenic spot objects. u ;

[0097] Secondly, from X u The problem of selecting k scenic spots is converted into a multi-objective optimization selection problem Maximize F(x) = (f1(x), f2(x), f3(x)), where f1(x), f2(x), and f3(x) are three different objective functions:

[0098] f1(x) represents the user’s interest in attraction x. i∈t,x i ∈S u , where sim(x,x i ) represents the recommended scenic spot x and the i-th scenic spot x visited by user u i The similarity of U x represents the set of users who visit attraction x, Indicates visiting attractions x i The user set of |U x| represents the total number of people visiting attraction x, Indicates visiting attractions x i Total number of people;

[0099] f2(x) represents the popularity of attraction x, v x represents the score of scenic spot x, f2(x) is normalized by min-max;

[0100] f3(x) represents the time between scenic spot x and the last time the user visited scenic spot x t The inverse of the distance between D(x,x t ) represents the relationship between scenic spot x and scenic spot x t The distance between them is normalized by min-max.

[0101] Furthermore, the specific steps of the to-be-optimized subset construction module constructing the to-be-optimized total set into multiple to-be-optimized subsets of travel recommendation solutions include:

[0102] First, count the total set X to be optimized u The number of times each attraction x is repeated;

[0103] Secondly, divide all scenic spots with a repetition number of 1 into subsets to be optimized Repeat this step until all scenic spots with a repetition number of z are divided into subsets to be optimized. A total of z subsets to be optimized are obtained, and the z subsets to be optimized are constructed into a set

[0104] In a preferred embodiment of the present invention, the specific calculation steps of the multi-objective optimization calculation module include:

[0105] First, for any subset i∈{1,2...z}, express A scenic spot in The number of scenic spots in the map is sorted by non-dominated sorting method. The calculation formula is as follows: for any x1 and x2, if the relationship in formula (5) is satisfied, it means that x2 dominates x1, f i 、f j All of them are objective functions, F represents the objective function set, and n Pareto fronts are obtained through non-dominated sorting: Front = {Front1, Front2, ..., Front n}, Front i represents the i-th Pareto frontier, and n represents the number of Pareto frontiers;

[0106]

[0107] Secondly, for any of the above Pareto fronts Front i ∈Front,i∈{1,2,...,n}, l≤j, for Front i For each individual x in the matrix, let the crowding distance of the border individual be infinite, and calculate the crowding distance of other individuals using formula (6), where x d The initial value of is 0, x+1 and x-1 represent the two individuals before and after x, and f i represents the objective function, |F| represents the total number of objective functions, and finally the Front i Sort each individual in ;

[0108]

[0109] Finally, combining the above two steps to calculate Sorting all individuals in the set to obtain the optimized subset.

[0110] Further, the final travel recommendation result output module calculates The internal individual sorting of all subsets to be optimized in Each optimized subset in Sort in order, and finally get The sorted set of all individuals after optimization in X′ is the total set after optimization, which is X′x. u The first k scenic spots are selected as the final output results of the model.

[0111] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A multi-model fusion tourism recommendation method based on a multi-objective optimization selection strategy, characterized in that: The following steps are involved: S1: Select two or more single recommendation models in the tourism recommendation problem, obtain the recommendation results of each single recommendation model, and merge all the results to obtain the total set to be optimized; S2: Model the multi-model fusion recommendation problem as a multi-objective optimization selection problem; the specific steps of step S2 include: S2.1: Based on the historical access records of user u , Indicates user i Attractions visited, select The single recommendation models predict the scenic spots visited by user u for the t+1th time, and obtain a set of z recommendation sequences with a length of k, which is recorded as , represents the recommendation result of the i-th single recommendation model, where , represents the jth scenic spot recommended by the i-th single recommendation model to user u, , , merge the results of each single recommendation model in L to construct a recommendation set containing z*k scenic spots ; S2.2: Will be The problem of selecting k scenic spots is converted into a multi-objective optimization selection problem , There are three different objective functions: represents the user's interest in attraction x, , , ,in Represents the recommended attraction x and the i-th attraction visited by user u The similarity of , represents the set of users who visit attraction x, Indicates visiting attractions A collection of users, represents the total number of people visiting attraction x, Indicates visiting attractions Total number of people; represents the popularity of attraction x, , represents the rating of attraction x, Perform min-max normalization processing; Indicates the time between attraction x and the last time the user visited the attraction The inverse of the distance between , Indicates attractions x and attractions The distance between them is normalized by min-max. S3: constructing the total set to be optimized into multiple sub-sets to be optimized of travel recommendation solutions; S4: Use the multi-objective optimization method to calculate the ranking of individuals in all subsets to be optimized and obtain the optimized subsets; S5: All optimized sub-sets are merged in the reverse order of the sorting of the sub-sets to be optimized to obtain the optimized total set. The first k attractions are selected from the optimized total set as the final output result of the model according to the needs, that is, the final tourism recommendation result is obtained.

2. The multi-model fusion tourism recommendation method based on multi-objective optimization selection strategy according to claim 1 is characterized in that: The specific steps of step S3 include: S3.1: Statistics of the total set to be optimized The number of times each attraction x is repeated; S3.2: Divide all scenic spots with a repetition count of 1 into subsets to be optimized Repeat this step until all scenic spots with a repetition number of z are divided into subsets to be optimized , a total of z subsets to be optimized are obtained, and the z subsets to be optimized are constructed into a set .

3. The multi-model fusion tourism recommendation method based on multi-objective optimization selection strategy according to claim 2 is characterized in that: The specific steps of step S4 include: S4.1: For any subset in step S3.2 , , , express A scenic spot in The number of scenic spots in the map is sorted by non-dominated sorting method. The calculation formula is as follows: , , if the relationship of formula (1) is satisfied, then it means Dominate , are the objective functions, Represents the objective function set, and a total of n Pareto frontiers are obtained through non-dominated sorting , represents the i-th Pareto frontier, and n represents the number of Pareto frontiers; （1）， S4.2: For any Pareto front in S4.1 , , , ,for For each individual x in the matrix, let the crowding distance of the border individual be infinite, and calculate the crowding distance of other individuals using formula (2), where The initial value of is 0. and represents the two individuals before and after x, represents the objective function, Represents the total number of objective functions, and finally sorts them according to the crowding distance from large to small Sort each individual in ; （2）， S4.3: Combine steps S4.1 and S4.2 to calculate the subset Sort all individuals in to get the optimized subset.

4. The multi-model fusion tourism recommendation method based on multi-objective optimization selection strategy according to claim 3 is characterized in that: The specific steps of step S5 include: S5.1: Calculate according to step S4 The internal individual sorting of all subsets to be optimized in Each optimized subset of Sort in order, and finally get The sorted set of all individuals after optimization, that is, the total set after optimization ; S5.2: According to the needs The first k scenic spots are selected as the final output results of the model.

5. A multi-model fusion tourism recommendation system based on multi-objective optimization selection strategy, characterized in that: include: The multi-model recommendation result calculation module is used to select two or more single recommendation models in the tourism recommendation problem, obtain the recommendation results of each single recommendation model, and merge all the results to obtain the total set to be optimized; Multi-objective optimization model building module, used to model the multi-model fusion recommendation problem as a multi-objective optimization selection problem; The specific steps of establishing the module modeling of the multi-objective optimization model include: First, based on the historical access records of user u , Indicates user i Attractions visited, select The single recommendation models predict the scenic spots visited by user u for the t+1th time, and obtain a set of z recommendation sequences with a length of k, which is recorded as , represents the recommendation result of the i-th single recommendation model, where , represents the jth scenic spot recommended by the i-th single recommendation model to user u, , , merge the results of each single recommendation model in L to construct a recommendation set containing z*k scenic spots ; Secondly, from The problem of selecting k scenic spots is converted into a multi-objective optimization selection problem , There are three different objective functions: represents the user's interest in attraction x, , , ,in Represents the recommended attraction x and the i-th attraction visited by user u The similarity of , represents the set of users who visit attraction x, Indicates visiting attractions A collection of users, represents the total number of people visiting attraction x, Indicates visiting attractions Total number of people; represents the popularity of attraction x, , represents the rating of attraction x, Perform min-max normalization processing; Indicates the time between attraction x and the last time the user visited the attraction The inverse of the distance between , Indicates attractions x and attractions The distance between them is normalized by min-max. A module for constructing sub-sets to be optimized, used for constructing the total set to be optimized into multiple sub-sets to be optimized of travel recommendation solutions; A multi-objective optimization calculation module is used to calculate the ranking of individuals in all subsets to be optimized using a multi-objective optimization method to obtain the optimized subsets; The final tourism recommendation result output module is used to merge all optimized sub-sets in the reverse order of the sorting of the sub-sets to be optimized to obtain the optimized total set, and select the first k attractions in the optimized total set as the final output result of the model according to the needs, that is, to obtain the final tourism recommendation result.

6. The multi-model fusion tourism recommendation system based on multi-objective optimization selection strategy according to claim 5 is characterized in that: The specific steps of the to-be-optimized subset construction module constructing the to-be-optimized total set into multiple to-be-optimized subsets of travel recommendation solutions include: First, count the total set to be optimized The number of times each attraction x is repeated; Secondly, divide all scenic spots with a repetition number of 1 into subsets to be optimized Repeat this step until all scenic spots with a repetition number of z are divided into subsets to be optimized , a total of z subsets to be optimized are obtained, and the z subsets to be optimized are constructed into a set .

7. The multi-model fusion tourism recommendation system based on multi-objective optimization selection strategy according to claim 6 is characterized in that: The specific calculation steps of the multi-objective optimization calculation module include: First, for any subset , , , express A scenic spot in The number of scenic spots in the map is sorted by non-dominated sorting method. The calculation formula is as follows: , , if the relationship of formula (1) is satisfied, then it means Dominate , are the objective functions, Represents the objective function set, and a total of n Pareto frontiers are obtained through non-dominated sorting , represents the i-th Pareto frontier, and n represents the number of Pareto frontiers; （1） ， Secondly, for any of the above Pareto frontiers , , , ,for For each individual x in the matrix, let the crowding distance of the border individual be infinite, and calculate the crowding distance of other individuals using formula (2), where The initial value of is 0. and represents the two individuals before and after x, represents the objective function, Represents the total number of objective functions, and finally sorts them according to the crowding distance from large to small Sort each individual in ; （2） ， Finally, combine the above two steps to calculate the subset Sort all individuals in to get the optimized subset.

8. The multi-model fusion tourism recommendation system based on multi-objective optimization selection strategy according to claim 7 is characterized in that: The final travel recommendation result output module is based on the calculated The internal individual sorting of all subsets to be optimized in Each optimized subset in Sort in order, and finally get The sorted set of all individuals after optimization, that is, the total set after optimization , according to the needs The first k scenic spots are selected as the final output results of the model.

Images