Personalized recommendation method and system for cruise meals based on zero historical data

Abstract

The present invention provides a cruise meal personalized recommendation method and system based on zero historical data, belonging to the field of cruise operation and maintenance services. The present invention aims at passengers who have never had cruise travel experience, extracts user meal preferences through questionnaire surveys, quantifies various indicators, then clusters the quantified preferences with cruise meals, and outputs clusters containing target users as inputs for collaborative filtering. In addition to considering meal preferences, the novelty of cruise passengers is also considered, and the novelty formula is used for calculation to obtain the final recommendation result. The present invention makes personalized recommendations to users based on user characteristics or meal characteristics, not only can the special meals of cruise restaurants be found, but also dishes that users may like can be recommended to users based on user preferences, which plays a relatively important role in improving the food experience of cruise passengers and has strong operability and applicability.

Description

Technical Field

[0001] The present invention belongs to the field of cruise ship sailing services, and in particular relates to a personalized cruise ship meal recommendation method based on zero historical data. Background Art

[0002] Cruise travel has the attributes of both sightseeing and transportation. At the same time, cruise travel has the characteristics of large passenger capacity, complex internal facilities, and a variety of services. From the perspective of cruise meals, the existing recommendations can only be recommended to users based on the special meals of each cruise restaurant, and lack personalized recommendations for each passenger. At the same time, for zero-history users who have never had a cruise travel experience, in the process of boarding a new cruise ship to start their trip, there is a cold start problem in terms of meals because the lack of historical data prevents users from interacting with cruise meals. If personalized recommendations can be made to users based on user characteristics or meal characteristics, not only can the special meals of cruise restaurants be found, but also dishes that users may like can be recommended to users based on user preferences, which plays a relatively important role in improving the food experience of cruise passengers.

[0003] The existing technology uses content-based collaborative filtering algorithms to recommend AR content to users or proposes a multi-task multi-tower model for personalized recommendation based on distillation. However, the former requires interactive behavior of content filtering, does not consider the need for fresh experience, and is not suitable for cruise meal recommendations; the latter requires a large amount of historical click data, which is not suitable for cruise ship scenes that lack historical data. Therefore, further research is needed on personalized recommendation methods with zero historical data that can be directly applied to cruise meals. Summary of the invention

[0004] In order to solve the problem of personalized recommendation of cruise ship meals in the prior art in the absence of historical data and improve the service level of cruise ship operation, the present invention provides a personalized recommendation method for cruise ship meals based on zero historical data.

[0005] In a first aspect of the present invention, the present invention provides a personalized cruise meal recommendation method based on zero historical data, comprising:

[0006] Step S1, extracting target user's meal preference features;

[0007] Step S2, obtaining cruise meal clusters similar to the target preference based on cluster analysis;

[0008] Step S3, calculating meal recommendation results taking into account the need for fresh experience.

[0009] Furthermore, the target user's meal preference feature extraction in step S1 specifically includes:

[0010] A questionnaire was designed, which included a variety of common meals for target users to choose from. Based on the target users' selection results, the target users' preference quantitative values ​​for "taste" T, "mouthfeel" H, and "cooking method" W were extracted; among them, "taste" was divided into five dimensions: sour, sweet, bitter, spicy, and salty, and each dimension was assigned a unique score of 1 to 9 points according to the degree of each dimension; "mouthfeel" was assigned a unique score of 1 to 9 points according to the degree of softness and hardness; "cooking method" was divided into nine methods: steaming, boiling, stewing, braising, roasting, stir-frying, frying, deep-frying, and smoking according to the degree of change in the characteristics of the ingredients, and each was assigned a unique score of 1 to 9 points.

[0011] Furthermore, in step S2, obtaining a cruise meal cluster similar to the target preference based on cluster analysis specifically includes:

[0012] Calculate the "distance" between the target user's preference characteristics and the cruise meal characteristics;

[0013] According to the distance calculation results, the K-means clustering method is used to cluster the cruise meals, and the cluster center closest to the target user's preference characteristics is determined. The meal set corresponding to the center is regarded as the cruise meal cluster similar to the target preference, denoted by K.

[0014] Furthermore, the “distance” between the target user’s preference characteristics and the cruise meal characteristics is calculated, including:

[0015] The "taste" distance R represents the taste similarity, and the specific calculation formula is:

[0016] ;

[0017] Among them, M represents all the data in the data set. , Represent the taste score of the target user’s preference features and the taste score of cruise meal j, respectively. represents the mean score of the calculated taste;

[0018] The "taste" distance F represents the taste similarity, and the specific calculation formula is:

[0019] ;

[0020] Among them, P represents the taste score in the target user's preference characteristics. represents the taste score of cruise ship meal j;

[0021] The “cooking method” distance N represents the similarity of cooking methods. The specific calculation formula is:

[0022] ;

[0023] Among them, A represents the cooking method score in the target user's preference characteristics, Represents the cooking style rating of cruise ship meal j.

[0024] Furthermore, step S3 calculates meal recommendation results taking into account the need for fresh experience, specifically including:

[0025] Calculate the cruise meal scores taking into account the need for novelty experience; based on the score calculation results, sort the meals in the cruise meal cluster that are similar to the target preference and give the recommendation results.

[0026] Furthermore, the cruise meal score calculation considering the need for fresh experience is calculated as follows:

[0027] ;

[0028] in, Represents the recommended rating of cruise ship meals. is the user's freshness weight coefficient for cruise meal j, Represents the taste similarity between meal j in the cruise meal cluster and the target user’s preference characteristics, Represents the taste similarity between meal j in the cruise meal cluster and the target user’s preference characteristics, Represents the similarity of cooking methods between meal j in the cruise meal cluster and the target user’s preference characteristics.

[0029] Furthermore, the user's freshness weight coefficient for cruise meal j is , the specific calculation method is:

[0030] Quantify the freshness characteristics of cruise ship meals according to target users;

[0031] The freshness weight coefficient reflecting the correlation between the user and the food is obtained through set operation.

[0032] Furthermore, the freshness characteristics of cruise ship meals are quantified according to the target users, including:

[0033] In the designed questionnaire, the target users are required to select the continent and climate environment where they live, and the corresponding quantitative values ​​are extracted according to their choices; the continent of residence includes six dimensions: Asia, Europe, Africa, North America, South America, and Oceania; the climate environment includes eleven dimensions: tropical rainforest, tropical grassland, tropical desert, tropical monsoon, Mediterranean, subtropical humid, temperate ocean, temperate continent, temperate monsoon, subarctic coniferous forest, and plateau mountain climate; thus, the user's feature set is obtained :

[0034] ;

[0035] in, Indicates the continent value of the user, represents the climate value of the user's living environment; similarly, for cruise meal j, its feature set is given :

[0036] ;

[0037] in, Indicates the continent to which the cruise meal belongs. Indicates the climate of the cruise ship meal.

[0038] Furthermore, the freshness weight coefficient of the relevance between the target user and the food is obtained through set operation, including:

[0039] Perform an intersection operation on the user's feature set and the feature set of cruise meal j to calculate the user's freshness weight coefficient for cruise meal j :

[0040] ;

[0041] in, Represents the function for calculating the number of elements in a collection. represents the set intersection operation, Represents an exponential function.

[0042] In a second aspect of the present invention, the present invention provides a cruise meal personalized recommendation system with zero historical data for implementing any method of the first aspect, comprising:

[0043] Feature extraction module, used to extract target user's meal preference features;

[0044] A calculation module is used to obtain cruise meal clusters similar to the target preference based on cluster analysis;

[0045] The recommendation module is used to calculate meal recommendation results based on the freshness experience requirements.

[0046] In a third aspect of the present invention, the present invention provides a computer device comprising a memory, a processor, and program instructions stored in the memory and executable by the processor, wherein the processor executes the program instructions to implement any one of the methods of the first aspect.

[0047] In a fourth aspect of the present invention, the present invention provides a computer-readable storage medium storing a computer program, wherein the computer program implements any one of the methods described in the first aspect when executed by a processor.

[0048] The present invention provides a cruise meal personalized recommendation method and system based on zero historical data, belonging to the field of cruise operation and maintenance services. The present invention aims at passengers who have never had cruise travel experience, extracts user meal preferences through questionnaire surveys, quantifies various indicators, then clusters the quantified preferences with cruise meals, and outputs clusters containing target users as inputs for collaborative filtering. In addition to considering meal preferences, the novelty of cruise passengers is also considered, and the novelty formula is used for calculation to obtain the final recommendation result. The present invention makes personalized recommendations to users based on user characteristics or meal characteristics, not only can the special meals of cruise restaurants be found, but also dishes that users may like can be recommended to users based on user preferences, which plays a relatively important role in improving the food experience of cruise passengers and has strong operability and applicability. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 —Schematic diagram of the overall process of the method of the present invention. DETAILED DESCRIPTION

[0050] The following will be combined with the drawings in the embodiments of the present application to further clearly and completely describe the technical solutions in the embodiments of the present application. It should be noted that the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without making creative work are within the scope of protection of this application.

[0051] In order to make the invention purpose, technical scheme and advantages of this application clearer, the technical scheme of the present invention is described clearly and completely in conjunction with the accompanying drawings and specific implementation schemes. It should be pointed out that for ordinary technicians in this field, several changes and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of this application.

[0052] like Figure 1 The figure shows an overall flow chart of a personalized recommendation method based on zero historical data. In the first embodiment, the present invention provides a personalized recommendation method for cruise meals based on zero historical data, including:

[0053] Step S1: Extracting target user’s meal preference features;

[0054] Step S2: Obtain cruise meal clusters similar to the target preference based on cluster analysis;

[0055] Step S3, calculating meal recommendation results taking into account the need for fresh experience.

[0056] In step S1, the characteristics of cruise passengers are extracted through a specific questionnaire survey, and the quantifiable issues are quantified. The designed questionnaire contains a variety of common meals for target users to choose from. According to the target user's selection results, the target user's preference quantitative values ​​for "taste" (expressed as T), "mouthfeel" (expressed as H), and "cooking method" (expressed as W) are extracted; among them, "taste" is divided into five dimensions: sour, sweet, bitter, spicy, and salty, and each dimension is assigned a unique score of 1 to 9 points; "mouthfeel" is assigned a unique score of 1 to 9 points according to the degree of hardness; "cooking method" is divided into nine methods: steaming, boiling, stewing, braising, roasting, stir-frying, frying, deep-frying, and smoking according to the degree of change in the characteristics of the ingredients, and each is assigned a unique score of 1 to 9 points.

[0057] Through the simulated questionnaire survey, we can get the quantitative score set of target user A:

[0058] {Corresponding scores for sour, sweet, bitter, spicy, salty, softness, hardness, and cooking methods|3, 7, 0, 0, 1, 6, 7};

[0059] In step S2, first, in the process of clustering using the K-means algorithm, since the cluster centers of the K-means algorithm are mostly randomly selected, the clustering effect cannot be measured during the calculation process. Therefore, it is necessary to set the K value range and find the optimal K value through iterative calculation.

[0060] In the iterative process, the most suitable K value is found by taking the value of the silhouette coefficient under different K values. In the test process, the silhouette coefficient values ​​of different k values ​​are observed from the range of K∈[2,9]. By calculating the silhouette coefficient, it is known that the entire silhouette coefficient fluctuates between 0.5-0.7. In general, it is believed that the clustering effect is better when S≥0.5. When k=6, S=0.71. Therefore, K=6 is the best clustering number.

[0061] Secondly, since different flavors represent different eating habits and preferences, it means that all flavors have "directionality" in the clustering process. Therefore, in the process of calculating different food flavors, cosine similarity is used for calculation. However, the biggest problem with traditional cosine similarity is that it only considers direction, and does not consider the impact of modulus size on similarity. Therefore, an improved cosine similarity method is used to calculate taste similarity. The formula is:

[0062] ;

[0063] Among them, R represents the taste similarity, M represents all the data in the data set, , Represent the taste of any two meals in the data set, Represents the average of the calculated flavors. By subtracting the average flavor of any meal, the similarity calculation can be reduced due to the modulus problem.

[0064] In terms of cooking methods, the nutritional value of meals can be judged by the different degrees of nutrient retention in cooking methods. In this dimension, nutritional similarity can be expressed as:

[0065] ;

[0066] Among them, N represents the similarity of cooking methods, A represents the cooking method score in the target user's preference characteristics, Represents the cooking style rating of cruise ship meal j.

[0067] In terms of taste, the taste is arranged from soft to hard. Since the taste has only one dimension, there is only a difference in "distance" in the dimension. In this dimension, the taste similarity can be expressed as:

[0068] ;

[0069] Among them, F represents the taste similarity, P represents the taste score in the target user's preference characteristics, Represents the taste score of cruise ship meals.

[0070] After K-means clustering, the cluster center closest to the target user A's preference characteristics is determined. The meal set corresponding to this center is the cruise meal cluster similar to the target preference, denoted by K:

[0071] {Meal number|Meal 27, Meal 8, Meal 15, Meal 25, Meal 13, Meal 7, Meal 38, Meal 4, Meal 22, Meal 30}

[0072] In step S3, meal recommendation results are calculated taking into account the need for fresh experience.

[0073] After clustering, clusters that meet the passengers' meal preferences have been obtained. It is still necessary to further consider sorting the meals according to the cruise passengers' sense of freshness. In the process of scoring calculation, in addition to considering the distance between user characteristics and meal characteristics, the impact of the target user's sense of freshness on meal recommendations should also be considered. Therefore, the following scoring calculation formula considering freshness is given:

[0074] ;

[0075] in, Represents the recommended rating of cruise ship meals. is the user's freshness weight coefficient for cruise meal j, Represents the taste similarity between meal j in the cruise meal cluster and the target user’s preference characteristics, Represents the taste similarity between meal j in the cruise meal cluster and the target user’s preference characteristics, represents the cooking method similarity between meal j in the cruise meal cluster and the target user’s preference characteristics. Next, we give Calculation method;

[0076] In the designed questionnaire, the target users are required to select the continent and climate environment where they live, and the corresponding quantitative values ​​are extracted according to their choices; the continent of residence includes six dimensions: Asia, Europe, Africa, North America, South America, and Oceania; the climate environment includes eleven dimensions: tropical rainforest, tropical grassland, tropical desert, tropical monsoon, Mediterranean, subtropical humid, temperate ocean, temperate continent, temperate monsoon, subarctic coniferous forest, and plateau mountain climate; thus, the user's feature set is obtained :

[0077] ;

[0078] in, Indicates the continent value of the user, represents the climate value of the user's living environment; similarly, for cruise meal j, its feature set is given :

[0079] ;

[0080] in, Indicates the continent to which the cruise meal belongs. Indicates the climate to which the cruise ship meal belongs;

[0081] Perform an intersection operation on the user's feature set and the feature set of cruise meal j to calculate the user's freshness weight coefficient for cruise meal j :

[0082] ;

[0083] in, Represents the function for calculating the number of elements in a collection. represents the set intersection operation, represents the exponential function;

[0084] Finally, using Sort the cruise meals by size and get meal recommendation results.

[0085] User A's recommended candidate list

[0086] Serial number Meal Name area climate acid sweet bitter hot salty Taste Cooking method Overall Rating 1 Meal 8 Europe Subtropical humid 3 5 0 0 0 5 6 0.84 2 Meal 4 Europe Temperate monsoon 3 3 0 0 0 7 9 0.96 3 Meal 7 Europe Temperate monsoon 0 5 0 0 1 5 4 1.39 4 Meal 38 Oceania Mediterranean 0 3 0 0 0 6 9 1.61 5 Meal 30 Europe Subtropical humid 0 7 0 0 0 5 8 1.67 6 Meal 25 Asia Temperate continents 1 5 0 0 0 5 7 2.30 7 Meal 22 Asia Tropical monsoon 0 5 0 0 1 3 5 2.34 8 Meal 13 North America Temperate Ocean 0 3 1 0 0 4 6 2.56 9 Meal 27 Asia Tropical monsoon 1 3 0 0 0 2 9 2.60 10 Meal 15 South America Subtropical humid 0 1 0 0 0 4 6 3.42

[0087] In the second embodiment of the present invention, the present invention provides a personalized cruise meal recommendation system with zero historical data for implementing any method in the first embodiment, including:

[0088] Feature extraction module, used to extract target user's meal preference features;

[0089] A calculation module is used to obtain cruise meal clusters similar to the target preference based on cluster analysis;

[0090] The recommendation module is used to calculate meal recommendation results based on the freshness experience requirements.

[0091] In a third embodiment of the present invention, the present invention provides a computer device, which includes a memory, a processor, and program instructions stored in the memory and executable by the processor, wherein the processor executes the program instructions to implement any one of the methods in the first embodiment.

[0092] In a fourth embodiment of the present invention, the present invention provides a computer-readable storage medium storing a computer program, wherein the computer program implements any one of the methods described in the first embodiment when executed by a processor.

[0093] In summary, the present invention provides a method for recommending meals for passengers who are boarding a newly commissioned cruise ship for the first time, and proposes a recommendation algorithm based on item clustering. The characteristics of the meals that the target user likes are determined through a questionnaire survey. After data processing, the data set is clustered, and the best number of clusters is selected through the silhouette coefficient value; the meals that are similar to the target user's historical data after clustering are used as input to calculate the similarity; the results are re-sorted according to the freshness to obtain the final recommendation. The present invention makes personalized recommendations to users based on user characteristics or meal characteristics, which can not only find the special meals of the cruise restaurant, but also recommend dishes that the user may like based on the user's preferences. It plays a relatively important role in improving the food experience of cruise passengers and has strong operability and applicability.

Claims

1. A personalized cruise meal recommendation method based on zero historical data, characterized in that: include: Step S1, extracting target user's meal preference features; Step S2, obtaining cruise meal clusters similar to the target preference based on cluster analysis; The method of obtaining a cruise meal cluster similar to the target preference based on cluster analysis specifically includes: Calculate the "distance" between the target user's preference characteristics and the cruise meal characteristics; According to the distance calculation results, the K-means clustering method is used to cluster the cruise meals, and the cluster center closest to the target user's preference characteristics is determined. The meal set corresponding to the center is regarded as the cruise meal cluster similar to the target preference, denoted as K; The calculation of the "distance" between the target user preference characteristics and the cruise meal characteristics specifically includes: The "taste" distance R represents the taste similarity, and the specific calculation formula is: ； Among them, M represents all the data in the data set. , Represent the taste score of the target user’s preference features and the taste score of cruise meal j, respectively. represents the mean score of the calculated taste; The "taste" distance F represents the taste similarity, and the specific calculation formula is: ； Among them, P represents the taste score in the target user's preference characteristics. represents the taste score of cruise ship meal j; The "cooking method" distance N represents the similarity of cooking methods. The specific calculation formula is: ； Among them, A represents the cooking method score in the target user's preference characteristics, represents the cooking method score of cruise ship meal j; Step S3, calculating meal recommendation results by considering the need for fresh experience; Step S3, calculating meal recommendation results by considering the need for fresh experience, specifically includes: Calculation of cruise meal scores considering the need for fresh experience; The specific calculation formula for the calculation of cruise meal scores considering the need for fresh experience is: ； in, Represents the recommended rating of cruise ship meals. is the user's freshness weight coefficient for cruise meal j, Represents the taste similarity between meal j in the cruise meal cluster and the target user’s preference characteristics, Represents the taste similarity between meal j in the cruise meal cluster and the target user’s preference characteristics, Represents the similarity of the cooking method of meal j in the cruise meal cluster with the target user's preference characteristics; wherein the user's freshness weight coefficient for cruise meal j ,The specific calculation method is: quantify the freshness characteristics of cruise meals according to the target users; obtain the freshness weight coefficient reflecting the correlation between users and meals through set operations; According to the score calculation results, the meals in the cruise meal cluster that are similar to the target preference are sorted and the recommendation results are given.

2. The personalized cruise meal recommendation method based on zero historical data according to claim 1 is characterized in that: The target user's meal preference feature extraction in step S1 specifically includes: A questionnaire was designed, which included a variety of common meals for target users to choose from. According to the target users' selection results, the target users' preference quantitative values ​​for "taste" T, "mouthfeel" H, and "cooking method" W were extracted; among them, "taste" was divided into five dimensions: sour, sweet, bitter, spicy, and salty, and each dimension was assigned a unique score of 1 to 9 points according to the degree of each dimension; "mouthfeel" was assigned a unique score of 1 to 9 points according to the degree of softness and hardness; "cooking method" was divided into nine methods: steaming, boiling, stewing, braising, roasting, stir-frying, frying, deep-frying, and smoking according to the degree of change in the characteristics of the ingredients, and each was assigned a unique score of 1 to 9 points.

3. The method according to claim 1, characterized in that The quantification of the freshness characteristics of cruise ship meals according to target users specifically includes: In the designed questionnaire, the target users are required to select the continent and climate environment where they live, and the corresponding quantitative values ​​are extracted according to their choices; the continent of residence includes six dimensions: Asia, Europe, Africa, North America, South America, and Oceania; the climate environment includes eleven dimensions: tropical rainforest, tropical grassland, tropical desert, tropical monsoon, Mediterranean, subtropical humid, temperate ocean, temperate continent, temperate monsoon, subarctic coniferous forest, and plateau mountain climate; thus, the user's feature set is obtained : ； in, Indicates the continent value of the user, represents the climate value of the user's living environment; similarly, for cruise meal j, its feature set is given : ； in, Indicates the continent to which the cruise meal belongs. Indicates the climate of the cruise ship meal.

4. The method according to claim 3, characterized in that The freshness weight coefficient of the relevance between the target user and the food obtained by set operation specifically includes: Perform an intersection operation on the user's feature set and the feature set of cruise meal j to calculate the user's freshness weight coefficient for cruise meal j : ； in, Represents the function for calculating the number of elements in a collection. represents the set intersection operation, Represents an exponential function.

5. A cruise meal personalized recommendation system with zero historical data for implementing the method described in any one of claims 1 to 3, characterized in that: include: Feature extraction module, used to extract target user's meal preference features; A calculation module is used to obtain cruise meal clusters similar to the target preference based on cluster analysis; The recommendation module is used to calculate meal recommendation results based on the freshness experience requirements.

6. A computer device comprising a memory, a processor and program instructions stored in the memory and executable by the processor, characterized in that: The processor executes the program instructions to implement the method according to any one of claims 1 to 4.

7. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 4 is implemented.

Images