A method and system for recommending ice and snow attractions by integrating preferences and eliminating popularity bias

Abstract

A method and system for recommending ice and snow attractions by integrating preferences and eliminating popularity deviations, relates to the technical field of attraction recommendation, and is used to solve the problem of over-recommendation caused by popularity deviations in existing recommendation methods. The technical points of the present invention include: grouping tourists by gender, age, and occupation, and mapping them into numerical values ​​of [0-1]; calculating the similarity of tourists' multi-feature preferences according to the mapped values ​​of tourists' gender, age, and occupation; calculating the tourists' preference values ​​according to the similarity of tourists' multi-feature preferences and tourists' historical rating data; constructing a tourist-preference value matrix according to the standardized tourist preference values; training a matrix decomposition model based on tourists' multi-feature preferences according to the tourist-preference value matrix and tourists' historical rating data to obtain the trained tourist-preference value matrix; predicting and recommending new users according to the trained tourist-preference value matrix. The present invention effectively alleviates the unfavorable situation that low-popularity ice and snow attractions are difficult to be recommended to tourists.

Description

Technical Field

[0001] The present invention relates to the technical field of scenic spot recommendation, and in particular to a method and system for recommending ice and snow scenic spots by integrating preferences and eliminating popularity deviations. Background Art

[0002] Tourism is a powerful support for promoting economic development. It provides great help for increasing income while also creating a unique city image. Among them, the recommendation of ice and snow attractions is unique and plays an important role in the recommendation of many tourist attractions. However, the current recommendation method of ice and snow attractions has a popularity bias problem, that is, the recommendation method mostly recommends popular, i.e., highly popular, ice and snow attractions to tourists. Although this can further enhance the recognition of the recommended ice and snow attractions, it will lead to the emergence of a long-tail phenomenon in the long run. That is, popular ice and snow attractions are over-recommended, and it is easy to cause excessive crowds, while niche ice and snow attractions and low-popularity attractions are relatively unpopular, resulting in loss of benefits. Therefore, the prior art does not solve the following problems: 1) Considering the long-tail phenomenon caused by over-recommending highly popular ice and snow attractions, the appropriate recommendation strength of low-popularity ice and snow attractions is not given; 2) While ensuring that the recommendation of ice and snow attractions has a high recommendation accuracy, the recommendation results meet the recommendation fairness and take into account the recommendation needs of niche and curious customers. Summary of the invention

[0003] In view of the above problems, the present invention proposes a method and system for recommending ice and snow attractions by integrating preferences and eliminating popularity deviation, so as to solve the problem of over-recommendation caused by popularity deviation in existing ice and snow attraction recommendation methods.

[0004] According to one aspect of the present invention, a method for recommending ice and snow attractions by integrating preferences and eliminating popularity deviations is provided, the method comprising the following steps:

[0005] Step 1: Obtain a training data set; the training data set includes historical tourist feature data and tourist historical rating data; the historical tourist feature data includes tourist ID, tourist gender, tourist age, and tourist occupation; the tourist historical rating data includes the ratings given by tourists to the ice and snow attractions they have visited;

[0006] Step 2: Group tourists by gender, age, and occupation, and map them into values ​​between [0 and 1];

[0007] Step 3: Calculate the tourist multi-feature preference similarity based on the mapping values ​​of the tourist gender, age, and occupation, where the tourist multi-feature preference similarity is used to represent the preference similarity of tourist groups with similar feature preferences;

[0008] Step 4: Calculate the tourist preference value based on the tourist multi-feature preference similarity and the tourist historical rating data;

[0009] Step 5: Standardize the tourist preference values, and make one-to-one correspondence between the standardized tourist preference values ​​and the tourist IDs and the ice and snow scenic spot IDs to construct a tourist-preference value matrix; the serial numbers of the rows and columns of the tourist-preference value matrix correspond to the tourist IDs and the ice and snow scenic spot IDs, respectively, and the values ​​of the tourist-preference value matrix are the standardized tourist preference values;

[0010] Step 6: training a matrix decomposition model based on tourists' multi-feature preferences according to the tourists' preference value matrix and the tourists' historical rating data, and obtaining a trained tourists' preference value matrix, which is defined as a tourists' multi-feature matrix;

[0011] Step 7: Input the ID of the tourist to be recommended or input the ID of the tourist to be recommended and the ID of the target ice and snow attraction, and search and obtain the predicted score values ​​of one or more ice and snow attractions according to the tourist-multi-feature matrix.

[0012] Furthermore, in step 2, linear mapping is used for tourist gender and tourist occupation, and trapezoidal fuzzy number mapping is used for tourist age. The trapezoidal fuzzy number formula is:

[0013]

[0014] Where x represents the input value of the tourist’s age; a,b,c,d∈R, and a≤b <c≤d;μ A represents the membership degree of trapezoidal fuzzy number, μ A ∈[0,1].

[0015] Furthermore, the calculation formula for the similarity of tourists' multi-feature preferences in step 3 is:

[0016]

[0017] Among them, TP(u,v) represents the preference similarity of tourists u and v; i,j 、v i,j They represent the corresponding mapping values ​​of tourists u and v when the input value of a certain feature j is i; when feature j is gender or occupation, k f =1, when feature j is age, k f represents the number of fuzzy sets; l represents the total number of features.

[0018] Furthermore, the calculation formula of the tourist preference value in step 4 is:

[0019]

[0020] Among them, P u,trepresents the tourist preference value of tourist u for attraction t; N represents the set of tourists with similar characteristic preferences; r v,t represents the actual rating value of tourist v for attraction t; They represent the average values ​​of all the scores given by tourists u and v to the ice and snow attractions they have visited.

[0021] Furthermore, in step 6, the original loss function of the matrix decomposition model is improved in the matrix decomposition model based on tourists' multi-feature preferences. The improved loss function is:

[0022]

[0023] Among them, k represents the dimension of latent factor space; r u,t represents the rating of tourist u on ice and snow attraction t; p u,k ,q k,t They represent the k-dimensional tourist latent factor matrix and the k-dimensional ice and snow attractions latent factor matrix respectively; λ represents the regularization coefficient; e is the natural index; n u,t represents the tourist preference value after standardization; M u,k represents the preference value of tourist u in dimension k; M k,t Represents the preference value of scenic spot t when the dimension is k.

[0024] Furthermore, in step 6, after obtaining the tourist-multi-feature matrix, score compensation is performed on the tourist-multi-feature matrix according to the standardized tourist preference value; wherein the score compensation calculation formula is:

[0025]

[0026] Among them, newM represents the tourist-multi-feature matrix after rating compensation; M u,t Represents the value in the tourist-preference value matrix of attraction t corresponding to tourist u.

[0027] According to another aspect of the present invention, a snow and ice scenic spot recommendation system is provided for integrating preferences and eliminating popularity deviations. The system comprises:

[0028] A data acquisition module is configured to acquire a training data set; the training data set includes historical tourist feature data and tourist historical rating data; the historical tourist feature data includes tourist ID, tourist gender, tourist age, and tourist occupation; the tourist historical rating data includes the ratings given by tourists to ice and snow attractions they have visited;

[0029] A grouping mapping module is configured to group tourists by gender, age, and occupation, and map the tourists' gender, age, and occupation into values ​​between [0 and 1];

[0030] The preference similarity calculation module is configured to calculate the tourist multi-feature preference similarity according to the mapping values ​​of the tourist gender, age, and occupation. The tourist multi-feature preference similarity is used to represent the preference similarity of tourist groups with similar feature preferences. The calculation formula of the tourist multi-feature preference similarity is:

[0031]

[0032] Among them, TP(u,v) represents the preference similarity of tourists u and v; i,j 、v i,j They represent the corresponding mapping values ​​of tourists u and v when the input value of a certain feature j is i; when feature j is gender or occupation, k f =1, when feature j is age, k f represents the number of fuzzy sets; l represents the total number of features;

[0033] The preference value calculation module is configured to calculate the tourist preference value according to the tourist multi-feature preference similarity and the tourist historical rating data; the calculation formula of the tourist preference value is:

[0034]

[0035] Among them, P u,t represents the tourist preference value of tourist u for attraction t; N represents the set of tourists with similar characteristic preferences; r v,t represents the actual rating value of tourist v for attraction t; They represent the average values ​​of all the scores given by tourists u and v to the ice and snow attractions they have visited;

[0036] A preference value matrix construction module is configured to perform standardization processing on the tourist preference values, and to make a one-to-one correspondence between the standardized tourist preference values ​​and the tourist IDs and the ice and snow scenic spot IDs, so as to construct a tourist-preference value matrix; the serial numbers of the rows and columns of the tourist-preference value matrix correspond to the tourist IDs and the ice and snow scenic spot IDs, respectively, and the values ​​of the tourist-preference value matrix are the standardized tourist preference values;

[0037] A preference value matrix training module is configured to train a matrix decomposition model based on tourist multi-feature preferences according to the tourist-preference value matrix and the tourist historical rating data, and obtain a trained tourist-preference value matrix, which is defined as a tourist-multi-feature matrix;

[0038] The recommendation module is configured to input the ID of the tourist to be recommended or the ID of the tourist to be recommended and the ID of the target ice and snow attraction, and to search and obtain the score prediction value of one or more ice and snow attractions according to the tourist-multi-feature matrix.

[0039] Furthermore, in the group mapping module, linear mapping is used for the tourist gender and tourist occupation, and trapezoidal fuzzy number mapping is used for the tourist age. The trapezoidal fuzzy number formula is:

[0040]

[0041] Where x represents the input value of the tourist’s age; a,b,c,d∈R, and a≤b <c≤d;μ A represents the membership degree of trapezoidal fuzzy number, μ A ∈[0,1].

[0042] Furthermore, in the preference value matrix training module, the original loss function of the matrix decomposition model is improved in the matrix decomposition model based on tourists' multi-feature preferences, and the improved loss function is:

[0043]

[0044] Among them, k represents the dimension of latent factor space; r u,t represents the rating of tourist u on ice and snow attraction t; p u,k ,q k,t They represent the k-dimensional tourist latent factor matrix and the k-dimensional ice and snow attractions latent factor matrix respectively; λ represents the regularization coefficient; e is the natural index; n u,t represents the tourist preference value after standardization; M u,k represents the preference value of tourist u in dimension k; M k,t Represents the preference value of scenic spot t when the dimension is k.

[0045] Furthermore, after obtaining the tourist-multi-feature matrix, the preference value matrix training module performs score compensation on the tourist-multi-feature matrix according to the standardized tourist preference values; wherein the score compensation calculation formula is:

[0046]

[0047] Among them, newM represents the tourist-multi-feature matrix after rating compensation; M u,t Represents the value in the tourist-preference value matrix of attraction t corresponding to tourist u.

[0048] The beneficial technical effects of the present invention are:

[0049] The present invention groups tourists of the same type according to their gender, age and occupation information and maps the three characteristics of tourists into values ​​between [0 and 1] respectively, and further constructs a tourist-multi-feature matrix, so as to better capture the real preferences of tourists, rather than allowing the recommendation results to be affected by popularity deviation, thereby increasing the control of tourist preferences on recommendation results; and, through the historical rating information of tourists on ice and snow attractions, the tourist-multi-feature matrix is ​​integrated to construct a matrix decomposition model loss function based on tourists' multi-feature preferences, and tourists' multi-feature preferences are integrated into model training to alleviate the popularity deviation problem inherent in the matrix decomposition recommendation model; further, the standard preference value of tourists is used as a compensation item for the training result of the matrix decomposition model based on tourists' multi-feature preferences, so that the popularity deviation is greatly reduced, avoiding the problem of high-popularity ice and snow attractions being continuously over-recommended in the existing matrix decomposition recommendation model, and at the same time, effectively alleviating the unfavorable situation that low-popularity ice and snow attractions are difficult to be presented in the tourist recommendation list. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 The present invention is a schematic flowchart of a method for recommending ice and snow attractions by integrating preferences and eliminating popularity deviations according to an embodiment of the present invention.

[0051] Figure 2 It is a schematic structural diagram of a snow and ice scenic spot recommendation system that integrates preferences and eliminates popularity deviations according to an embodiment of the present invention. DETAILED DESCRIPTION

[0052] In order to enable those skilled in the art to better understand the scheme of the present invention, exemplary implementations or embodiments of the present invention will be described below in conjunction with the accompanying drawings. Obviously, the described implementations or embodiments are only implementations or embodiments of a part of the present invention, not all of them. Based on the implementations or embodiments of the present invention, all other implementations or embodiments obtained by ordinary technicians in the field without creative work should fall within the scope of protection of the present invention.

[0053] The present invention fully considers the problem that popular ice and snow attractions are affected by popularity deviation, resulting in over-recommendation. Over-recommendation of popular ice and snow attractions will lead to a long-tail phenomenon in the recommendation, causing traffic to gather at ice and snow attractions with high popularity, thereby leading to crowd gathering. On the other hand, due to the unfair recommendation caused by popularity deviation, low-popularity ice and snow attractions are difficult to be pushed to the tourist preference list, which is an unfair recommendation for both ice and snow attractions and tourists. In view of the above problems, the present invention combines tourists' multi-feature information to capture tourists' real preferences, reduce the impact of popularity deviation on the recommendation results of ice and snow attractions, and increase the influence of tourists' real preference values ​​on the recommendation results. At the same time, the present invention adopts different mapping inverse methods according to the properties of different characteristic information of tourists, further improves the accuracy of tourists' preference capture, and effectively reduces the adverse problems caused by popularity deviation; the present invention integrates tourists' multi-feature preferences into the matrix decomposition model to optimize the inherent popularity deviation of the traditional matrix decomposition model, and at the same time, the optimized preliminary recommendation results are scored and compensated again using tourists' multi-feature preference values, so as to further reduce the popularity deviation generated by the recommendation model based on matrix decomposition in its inherent design mechanism, and improve the fairness of recommendation.

[0054] A snow and ice attraction recommendation method that integrates preferences and eliminates popularity bias, such as Figure 1 As shown, the following steps are included:

[0055] Step 1: Obtain a training data set; the training data set includes historical tourist feature data and tourist historical rating data; the historical tourist feature data includes tourist ID, tourist gender, tourist age, and tourist occupation; the tourist historical rating data includes the ratings given by tourists to the ice and snow attractions they have visited;

[0056] Step 2: Group tourists by gender, age, and occupation, and map them into values ​​between [0 and 1];

[0057] Step 3: Calculate the similarity of tourists' multi-feature preferences based on the mapping values ​​of tourists' gender, age, and occupation. The similarity of tourists' multi-feature preferences is used to represent the preference similarity of tourist groups with similar feature preferences.

[0058] Step 4: Calculate the tourist preference value based on the tourist multi-feature preference similarity and the tourist historical rating data;

[0059] Step 5: Standardize the tourist preference values, and make a one-to-one correspondence between the standardized tourist preference values ​​and the tourist ID and the ice and snow attraction ID to construct a tourist-preference value matrix; the serial numbers of the rows and columns of the tourist-preference value matrix correspond to the tourist ID and the ice and snow attraction ID, respectively, and the values ​​of the tourist-preference value matrix are the standardized tourist preference values;

[0060] Step 6: Train a matrix decomposition model based on tourists' multi-feature preferences according to the tourists' preference value matrix and tourists' historical rating data to obtain a trained tourists' preference value matrix, which is defined as a tourists' multi-feature matrix.

[0061] Step 7: Enter the ID of the tourist to be recommended or the ID of the tourist to be recommended and the ID of the target ice and snow attraction, and search for the predicted score values ​​of one or more ice and snow attractions based on the tourist-multi-feature matrix.

[0062] An embodiment of the present invention provides a method for recommending ice and snow attractions by integrating preferences and eliminating popularity deviations. The method comprises the following steps:

[0063] Step 1: Obtain tourist characteristic information and tourist historical rating information; wherein the tourist characteristic information includes the tourist's gender, age, and occupation; the tourist historical rating information includes the tourist's rating scores for the ice and snow attractions that he has visited;

[0064] According to an embodiment of the present invention, historical rating information of tourists on ice and snow attractions and gender, age, and occupation in registration information are obtained, and preliminary grouping is performed based on the gender, age, and occupation of tourists, that is, the same gender / age / occupation is grouped together.

[0065] Step 2: According to the tourists’ gender information, the tourists are grouped according to their gender, and the groups include “male”, “female”, and “tourists with no set gender”; according to the tourists’ age information, the tourists are grouped according to their age using trapezoidal fuzzy numbers, and the groups include “youth”, “middle-aged”, and “elderly”; according to the tourists’ occupation information, the tourists are grouped according to their occupation, and the groups include “other”, “academic / educator”, “artist”, “clerk / administrator”, “college student / graduate student”, “customer service”, “doctor / health care”, “executive / management”, “farmer”, “housewife”, “primary / middle / high school student”, “lawyer”, “programmer”, “retired”, “sales / marketing”, “scientist”, “self-employed”, “technician / engineer”, “businessman / craftsman”, “unemployed”, and “writer”;

[0066] According to an embodiment of the present invention, gender is set as "male" = 1, "female" = 0, "tourist whose gender is not set" = 0.5 when input; the gender input value can also be mapped to "male" = 1, "female" = 0, "tourist whose gender is not set" = 0.5 using y1 = x1, where x1 is the gender input value and y1 is the gender mapping value.

[0067] According to an embodiment of the present invention, the age is input according to a traditional age value, such as 20 (years old); the age input value can also be mapped into a continuous value between [0 and 1] using a trapezoidal fuzzy number; the trapezoidal fuzzy number formula is:

[0068]

[0069] Among them, a,b,c,d∈R, and a≤b <c≤d;μ A is the membership degree of the trapezoidal fuzzy number, μ A ∈[0,1].

[0070] According to the embodiment of the present invention, occupations are inputted as values ​​between [0 and 20] in the order of grouping, for example, "other" = 0, "academic / educator" = 1; occupation input values ​​are in the form of Mapped into a continuous value between [0 and 1], where x2 is the occupation input value and y2 is the occupation mapping value.

[0071] The gender, age, and occupational characteristics of tourists are mapped to the interval [0-1]. Considering the fuzziness of tourists' age, that is, a certain age cannot be mapped to a certain age group, but can be mapped to one or more age groups, trapezoidal fuzzy numbers are used to map tourists' age. Fuzzy numbers are a concept in fuzzy mathematics. Fuzzy numbers have good performance in dealing with uncertain factors. Combined with the characteristics of tourists' age, trapezoidal fuzzy numbers are used to map tourists' age to the interval [0-1], which effectively solves the fuzziness of age. The mapping result is a continuous value in the interval [0-1].

[0072] Since the gender and occupation of tourists are not ambiguous and are deterministic features, when considering the gender of tourists, the input value of male is set to 1, the input value of female is set to 0, and the input value of tourists whose gender is not set is set to 0.5. And y1=x1 is used to map "male"=1, "female"=0, "tourist whose gender is not set"=0.5, where x1 is the gender input value and y1 is the gender mapping value.

[0073] When considering tourists' occupations, 21 occupations are included, including "others", "academic / educator", "artist", "clerk / administrator", "college student / graduate student", "customer service", "doctor / health care", "executive / management", "farmer", "housewife", "primary / middle / high school student", "lawyer", "programmer", "retired", "sales / marketing", "scientist", "self-employed", "technician / engineer", "businessman / craftsman", "unemployed", and "writer". They are converted into input values ​​[0-20] in sequence, that is, "others" = 0, "academic / educator" = 1, "artist" = 3, and so on. And adopt Mapped into a continuous value between [0 and 1], where x2 is the occupation input value and y2 is the occupation mapping value mapped into a continuous value between [0 and 1].

[0074] Based on the different properties of tourist characteristics, the present invention performs different mapping processes on different characteristics, avoiding the rigid division in traditional recommendation methods and making the characteristics with fuzzy properties tangible. The detailed information is shown in the following table:

[0075]

[0076]

[0077] Step 3: Use the tourist multi-feature preference similarity calculation formula to calculate the tourist groups with similar feature preferences. The tourist multi-feature preference similarity calculation formula is as follows:

[0078]

[0079] Among them, TP(u,v) represents the preference similarity of tourists u and v; i,j 、v i,j They represent the mapping results of tourists u and v when the input value of a certain feature j is i; when feature j is gender or occupation, k f =1, when feature j is age, k f represents the number of fuzzy sets; l represents the total number of features.

[0080] Step 4: Use the weighted average strategy calculation formula to calculate the tourist preference value based on the similarity of tourists' multi-feature preferences. The weighted average strategy calculation formula is as follows:

[0081]

[0082] Among them, P u,t represents the tourist preference value of tourist u for scenic spot t; N represents the set of tourists with similar characteristic preferences; They represent the average of all the scores given by tourists u and v to the ice and snow attractions they have visited; TP(u,v) represents the similarity of preferences between tourists u and v; r v,t represents the actual rating given by tourist v to attraction t.

[0083] The calculation results of tourists’ preference values ​​are shown in the following table:

[0084] Visitor ID Snow and ice attraction ID Preference value 1 20 3.01 2 13 3.48 4 50 5.20

[0085] Step 5: Use a standardized calculation formula to standardize the tourist preference values ​​in step 4, and construct a tourist-preference value matrix M based on the standardized tourist preference values; specifically: the standardized tourist preference values ​​are matched one-to-one with the tourist ID and the ice and snow scenic spot ID to construct a tourist-preference value matrix; the row and column numbers of the tourist-preference value matrix correspond to the tourist ID and the ice and snow scenic spot ID respectively, and the values ​​of the tourist-preference value matrix are the standardized tourist preference values;

[0086] According to an embodiment of the present invention, the standardized calculation formula is as follows:

[0087]

[0088] Among them, U represents the total number of tourists; C represents the total number of ice and snow attractions; P u,t represents the tourist preference value of tourist u for attraction t; n u,t Indicates that P u,t The normalized value.

[0089] Step 6: The obtained tourist-preference value matrix M and the historical rating information of tourists on ice and snow attractions are integrated to construct a matrix decomposition loss function based on tourists' multi-feature preferences, and the loss function is used as the objective function of the matrix decomposition model. The tourist-multi-feature matrix is ​​obtained by training the constructed matrix decomposition model;

[0090] According to an embodiment of the present invention, matrix decomposition plays a very important role in the recommendation model. Matrix decomposition uses tourists' explicit data (ratings / preference values, etc.) or implicit data (click or not / purchase or not, etc.) as raw data to construct a co-occurrence matrix, and then continuously iterates and updates the raw data to decompose it into potential factors of tourists and ice and snow attractions, and then further fills in the unknown values ​​in the matrix to achieve the purpose of predicting tourists' preferences for ice and snow attractions. The loss function of the original matrix decomposition model is as follows:

[0091]

[0092] The present invention improves the original loss function, and the matrix decomposition loss function based on tourists' multi-feature preferences is as follows:

[0093]

[0094] Among them, k represents the dimension of latent factor space; r u,t represents the rating of tourist u on ice and snow attraction t; p u,k ,q k,t They represent the k-dimensional tourist latent factor matrix and the k-dimensional ice and snow attractions latent factor matrix respectively; λ represents the regularization coefficient; e is the natural index; n u,t represents the tourist preference value after standardization; Mu,k represents the preference value of tourist u in dimension k; M k,t Represents the preference value of scenic spot t when the dimension is k.

[0095] Experimental verification shows that the matrix decomposition model based on tourists' multi-feature preferences is superior to the original matrix decomposition model in terms of recommendation accuracy and reducing popularity deviation.

[0096] The model training process is as follows:

[0097] First, the data obtained in the first step is divided into training set, validation set, and test set in a ratio of 7:2:1.

[0098] Secondly, the training set is used to train the parameters of the matrix decomposition model based on tourists' multi-feature preferences; the validation set is used to train the hyperparameters of the matrix decomposition model based on tourists' multi-feature preferences; and the test set is used to evaluate the prediction accuracy of the matrix decomposition model based on tourists' multi-feature preferences.

[0099] Finally, enter the target tourist ID to obtain a preliminary recommendation prediction score. Specifically, enter the tourist ID to be recommended or the tourist ID to be recommended and the target ice and snow attraction ID, and find and obtain the score prediction value of one or more ice and snow attractions based on the tourist-multi-feature matrix. Among them, when only the target tourist ID is entered, the row where the tourist ID is located is the current tourist's recommended prediction score for all ice and snow attractions; when the target tourist ID and the target attraction ID are entered, it indicates the tourist's recommended prediction score for the attraction. The preliminary recommendation prediction score is shown in the following table:

[0100]

[0101] Step 7: Based on the standardized tourist preference values ​​obtained in step 5, the tourist-multi-feature matrix described in step 6 is scored and compensated; and the final tourist score prediction value for the scenic spot is obtained;

[0102] According to an embodiment of the present invention, the scoring compensation formula is:

[0103]

[0104] Among them, newM is the matrix after score compensation; M u,t Represents the value in the tourist-preference value matrix of attraction t corresponding to tourist u.

[0105] The final predicted values ​​of tourists' ratings for ice and snow attractions are shown in the following table:

[0106]

[0107] Step 8: Sort the final score prediction values ​​obtained from large to small, and select the top (K) attractions to recommend to tourists, where K represents the number of recommendations.

[0108] According to an embodiment of the present invention, according to the predicted scores of tourists for ice and snow attractions, the top (K) results are selected from high to low and recommended to tourists, where K is the number of recommended attractions. The recommended examples are shown in the following table. After the above steps, the predicted scores of tourists for ice and snow attractions are obtained, and K=3 is set, indicating that the top three attractions with the highest predicted scores are recommended to tourists.

[0109]

[0110]

[0111] Based on the Top(3) recommendation, the final recommendation list given by the tourism system to tourists is as follows:

[0112]

[0113] Another embodiment of the present invention provides a snow and ice scenic spot recommendation system that integrates preferences and eliminates popularity deviations. Figure 2 As shown, the system includes:

[0114] The data acquisition module 10 is configured to acquire a training data set; the training data set includes historical tourist feature data and tourist historical rating data; the historical tourist feature data includes tourist ID, tourist gender, tourist age, and tourist occupation; the tourist historical rating data includes the ratings given by tourists to the ice and snow attractions they have visited;

[0115] A grouping mapping module 20 is configured to group tourists by gender, age, and occupation, and map the tourists' gender, age, and occupation into values ​​between [0 and 1];

[0116] The preference similarity calculation module 30 is configured to calculate the tourist multi-feature preference similarity according to the mapping values ​​of the tourist gender, age, and occupation. The tourist multi-feature preference similarity is used to represent the preference similarity of tourist groups with similar feature preferences. The calculation formula of the tourist multi-feature preference similarity is:

[0117]

[0118] Among them, TP(u,v) represents the preference similarity of tourists u and v; i,j 、v i,j They represent the corresponding mapping values ​​of tourists u and v when the input value of a certain feature j is i; when feature j is gender or occupation, k f =1, when feature j is age, k f represents the number of fuzzy sets; l represents the total number of features;

[0119] The preference value calculation module 40 is configured to calculate the tourist preference value according to the tourist multi-feature preference similarity and the tourist historical rating data; the calculation formula of the tourist preference value is:

[0120]

[0121] Among them, P u,t represents the tourist preference value of tourist u for attraction t; N represents the set of tourists with similar characteristic preferences; r v,t represents the actual rating value of tourist v for attraction t; They represent the average values ​​of all the scores given by tourists u and v to the ice and snow attractions they have visited;

[0122] The preference value matrix construction module 50 is configured to perform standardization processing on the tourist preference values, and to make a one-to-one correspondence between the standardized tourist preference values ​​and the tourist ID and the ice and snow attraction ID, so as to construct a tourist-preference value matrix; the serial numbers of the rows and columns of the tourist-preference value matrix correspond to the tourist ID and the ice and snow attraction ID, respectively, and the values ​​of the tourist-preference value matrix are the standardized tourist preference values;

[0123] A preference value matrix training module 60 is configured to train a matrix decomposition model based on tourist multi-feature preferences according to the tourist-preference value matrix and tourist historical rating data to obtain a trained tourist-preference value matrix, which is defined as a tourist-multi-feature matrix;

[0124] The recommendation module 70 is configured to input the ID of the tourist to be recommended or the ID of the tourist to be recommended and the ID of the target ice and snow attraction, and to search and obtain the score prediction value of one or more ice and snow attractions according to the tourist-multi-feature matrix.

[0125] In this embodiment, preferably, the group mapping module 20 uses linear mapping for the tourist gender and tourist occupation, and uses trapezoidal fuzzy number mapping for the tourist age. The trapezoidal fuzzy number formula is:

[0126]

[0127] Where x represents the input value of the tourist’s age; a,b,c,d∈R, and a≤b <c≤d;μ A represents the membership degree of trapezoidal fuzzy number, μ A ∈[0,1].

[0128] In this embodiment, preferably, the preference value matrix training module 60 improves the original loss function of the matrix decomposition model in the matrix decomposition model based on tourists' multi-feature preferences, and the improved loss function is:

[0129]

[0130] Among them, k represents the dimension of latent factor space; r u,t represents the rating of tourist u on ice and snow attraction t; p u,k ,q k,t They represent the k-dimensional tourist latent factor matrix and the k-dimensional ice and snow attractions latent factor matrix respectively; λ represents the regularization coefficient; e is the natural index; n u,t represents the tourist preference value after standardization; M u,k represents the preference value of tourist u in dimension k; M k,t Represents the preference value of scenic spot t when the dimension is k.

[0131] In this embodiment, preferably, after obtaining the tourist-multi-feature matrix, the preference value matrix training module 60 performs score compensation on the tourist-multi-feature matrix according to the standardized tourist preference value; wherein the score compensation calculation formula is:

[0132]

[0133] Among them, newM represents the tourist-multi-feature matrix after rating compensation; M u,t Represents the value in the tourist-preference value matrix of attraction t corresponding to tourist u.

[0134] The functions of a snow and ice attraction recommendation system that integrates preferences and eliminates popularity deviations in an embodiment of the present invention can be described by the aforementioned snow and ice attraction recommendation method that integrates preferences and eliminates popularity deviations. Therefore, for the parts not described in detail in the system embodiment, please refer to the above method embodiment and will not be repeated here.

[0135] Although the present invention has been described according to a limited number of embodiments, it will be apparent to those skilled in the art, with the benefit of the above description, that other embodiments are contemplated within the scope of the invention thus described. The disclosure of the present invention is intended to be illustrative rather than restrictive of the scope of the invention, which is defined by the appended claims.

Claims

1. A method for recommending ice and snow attractions by integrating preferences and eliminating popularity deviations, characterized in that: The following steps are involved: Step 1: Obtain a training data set; the training data set includes historical tourist feature data and tourist historical rating data; the historical tourist feature data includes tourist ID, tourist gender, tourist age, and tourist occupation; the tourist historical rating data includes the ratings given by tourists to the ice and snow attractions they have visited; Step 2: Group tourists by gender, age, and occupation, and map them into values ​​between [0 and 1]; Step 3: Calculate the tourist multi-feature preference similarity based on the mapping values ​​of the tourist gender, age, and occupation, where the tourist multi-feature preference similarity is used to represent the preference similarity of tourist groups with similar feature preferences; Step 4: Calculate the tourist preference value based on the tourist multi-feature preference similarity and the tourist historical rating data; Step 5: Standardize the tourist preference values, and make one-to-one correspondence between the standardized tourist preference values ​​and the tourist IDs and the ice and snow scenic spot IDs to construct a tourist-preference value matrix; the serial numbers of the rows and columns of the tourist-preference value matrix correspond to the tourist IDs and the ice and snow scenic spot IDs, respectively, and the values ​​of the tourist-preference value matrix are the standardized tourist preference values; Step 6: training a matrix decomposition model based on tourists' multi-feature preferences according to the tourists' preference value matrix and the tourists' historical rating data, and obtaining a trained tourists' preference value matrix, which is defined as a tourists' multi-feature matrix; Step 7: Input the ID of the tourist to be recommended or input the ID of the tourist to be recommended and the ID of the target ice and snow attraction, and search and obtain the predicted score values ​​of one or more ice and snow attractions according to the tourist-multi-feature matrix.

2. According to claim 1, a method for recommending ice and snow attractions by integrating preferences and eliminating popularity deviations is characterized in that: In step 2, linear mapping is used for tourist gender and tourist occupation, and trapezoidal fuzzy number mapping is used for tourist age. The trapezoidal fuzzy number formula is: Where x represents the input value of the tourist’s age; a, b, c, d ∈ R, and a ≤ b < c ≤ d; μ A represents the membership degree of trapezoidal fuzzy number, μ A ∈[0,1].

3. The method for recommending ice and snow attractions by integrating preferences and eliminating popularity deviation according to claim 2, characterized in that: The calculation formula for tourists’ multi-feature preference similarity in step 3 is: Among them, TP(u,v) represents the preference similarity of tourists u and v; i,j 、v i,j They represent the corresponding mapping values ​​of tourists u and v when the input value of a certain feature j is i; when feature j is gender or occupation, k f =1, when feature j is age, k f represents the number of fuzzy sets; l represents the total number of features.

4. The method for recommending ice and snow attractions by integrating preferences and eliminating popularity deviation according to claim 3, characterized in that: The calculation formula of the tourist preference value in step 4 is: Among them, P u,t represents the tourist preference value of tourist u for attraction t; N represents the set of tourists with similar characteristic preferences; r v,t represents the actual rating value of tourist v for attraction t; They represent the average values ​​of all the scores given by tourists u and v to the ice and snow attractions they have visited.

5. According to claim 4, a method for recommending ice and snow attractions by integrating preferences and eliminating popularity deviations is characterized in that: In step 6, the original loss function of the matrix decomposition model is improved in the matrix decomposition model based on tourists' multi-feature preferences. The improved loss function is: Among them, k represents the dimension of latent factor space; r u,t represents the rating of tourist u on ice and snow attraction t; p u,k ,q k,t They represent the k-dimensional tourist latent factor matrix and the k-dimensional ice and snow attractions latent factor matrix respectively; λ represents the regularization coefficient; e is the natural index; n u,t represents the tourist preference value after standardization; M u,k represents the preference value of tourist u in dimension k; M k,t Represents the preference value of scenic spot t when the dimension is k.

6. The method for recommending ice and snow attractions by integrating preferences and eliminating popularity deviation according to claim 5, characterized in that: Step 6: After obtaining the tourist-multi-feature matrix, score compensation is performed on the tourist-multi-feature matrix according to the standardized tourist preference value; wherein the score compensation calculation formula is: Among them, newM represents the tourist-multi-feature matrix after rating compensation; M u,t Represents the value in the tourist-preference value matrix of attraction t corresponding to tourist u.

7. A snow and ice scenic spot recommendation system that integrates preferences and eliminates popularity deviations, characterized in that: include: A data acquisition module is configured to acquire a training data set; the training data set includes historical tourist feature data and tourist historical rating data; the historical tourist feature data includes tourist ID, tourist gender, tourist age, and tourist occupation; the tourist historical rating data includes the ratings given by tourists to ice and snow attractions they have visited; A grouping mapping module is configured to group tourists by gender, age, and occupation, and map the tourists' gender, age, and occupation into values ​​between [0 and 1]; The preference similarity calculation module is configured to calculate the tourist multi-feature preference similarity according to the mapping values ​​of the tourist gender, age, and occupation. The tourist multi-feature preference similarity is used to represent the preference similarity of tourist groups with similar feature preferences. The calculation formula of the tourist multi-feature preference similarity is: Among them, TP(u,v) represents the preference similarity of tourists u and v; i,j 、v i,j They represent the corresponding mapping values ​​of tourists u and v when the input value of a certain feature j is i; when feature j is gender or occupation, k f =1, when feature j is age, k f represents the number of fuzzy sets; l represents the total number of features; The preference value calculation module is configured to calculate the tourist preference value according to the tourist multi-feature preference similarity and the tourist historical rating data; the calculation formula of the tourist preference value is: Among them, P u,t represents the tourist preference value of tourist u for attraction t; N represents the set of tourists with similar characteristic preferences; r v,t represents the actual rating value of tourist v for attraction t; They represent the average values ​​of all the scores given by tourists u and v to the ice and snow attractions they have visited; A preference value matrix construction module is configured to perform standardization processing on the tourist preference values, and to make a one-to-one correspondence between the standardized tourist preference values ​​and the tourist IDs and the ice and snow scenic spot IDs, so as to construct a tourist-preference value matrix; the serial numbers of the rows and columns of the tourist-preference value matrix correspond to the tourist IDs and the ice and snow scenic spot IDs, respectively, and the values ​​of the tourist-preference value matrix are the standardized tourist preference values; A preference value matrix training module is configured to train a matrix decomposition model based on tourist multi-feature preferences according to the tourist-preference value matrix and the tourist historical rating data, and obtain a trained tourist-preference value matrix, which is defined as a tourist-multi-feature matrix; The recommendation module is configured to input the ID of the tourist to be recommended or the ID of the tourist to be recommended and the ID of the target ice and snow attraction, and to search and obtain the score prediction value of one or more ice and snow attractions according to the tourist-multi-feature matrix.

8. The ice and snow attractions recommendation system for integrating preferences and eliminating popularity deviation according to claim 7 is characterized in that: In the group mapping module, linear mapping is used for the gender and occupation of tourists, and trapezoidal fuzzy number mapping is used for the age of tourists. The trapezoidal fuzzy number formula is: Where x represents the input value of the tourist’s age; a, b, c, d ∈ R, and a ≤ b < c ≤ d; μ A represents the membership degree of trapezoidal fuzzy number, μ A ∈[0,1].

9. The ice and snow attractions recommendation system for integrating preferences and eliminating popularity deviation according to claim 8 is characterized in that: In the preference value matrix training module, the original loss function of the matrix decomposition model is improved in the matrix decomposition model based on tourists' multi-feature preferences. The improved loss function is: Among them, k represents the dimension of latent factor space; r u,t represents the rating of tourist u on ice and snow attraction t; p u,k ,q k,t They represent the k-dimensional tourist latent factor matrix and the k-dimensional ice and snow attractions latent factor matrix respectively; λ represents the regularization coefficient; e is the natural index; n u,t represents the tourist preference value after standardization; M u,k represents the preference value of tourist u in dimension k; M k,t Represents the preference value of scenic spot t when the dimension is k.

10. The ice and snow attractions recommendation system for integrating preferences and eliminating popularity deviation according to claim 9 is characterized in that: After obtaining the tourist-multi-feature matrix, the preference value matrix training module performs score compensation on the tourist-multi-feature matrix according to the standardized tourist preference value; wherein the score compensation calculation formula is: Among them, newM represents the tourist-multi-feature matrix after rating compensation; M u,t Represents the value in the tourist-preference value matrix of attraction t corresponding to tourist u.

Images