Based On Users' Rating To Personalized Diet Recommendation Service

With the improvement of living standards,a variety of multi-flavored diet has become a topic that most people concern.Due to the limited knowledge of diet recipes,most people find information through search engines or gourmet food websites,choose recipes,and plan their diets.However,search engine's result has too much redundant information,and it's inconvenient to use.Diet recommendation system based on user information,preferences provides dietary recommendations.It's convenient and popular with more and more people of all ages.At present,most domestic food recommendation systems are based on gourmet websites and mobile phone APPs,and are recommended by users for providing personal information.However,the current diet recommendation services have the following problems and difficulties:First,the user's uncertainty about the preference of the recipe.Second,due to the sparseness of the user-recipes score matrix data,the accuracy of collaborative filtering recommendation users' scores is low.Third,the existing recipes recommendation do not consider the nutritional balance Problems,does not meet the modern concept of healthy diet.In view of the existing problems of diet recommendation methods,this thesis studies on the basis of previous studies.The contributions and innovations of this paper include the following points:(1)Adopting a method of constructing user preference model to conduct personalized recipes recommendation.Firstly,we select some characteristic attributes of the recipe to describe the recipe,including the cooking method,cooking difficulty,taste and so on.Then,based on the recipe characteristics and scores that the user has evaluated,we use the machine learning method to fit the user preference model.Finally,calculate the user's rating of unknown recipes and recommend the recipes of high ratings to the user.(2)Adopting collaborative filtering recommendation algorithm populated based on preference model for recipe recommendation.First,in order to solve the problem that the recommendation result of the preference model tends to be over-specialized and lacks the potential mining ability to users,this paper chooses the recommendation of collaborative filtering based on users for recipe recommendation.Then,in order to solve the problem of collaborative filtering recommendation,This thesis chooses the method of score filling,constructs the preference model for each user,predicts the score of the unrated item,calculates the user similarity and finds the nearest neighbor by using the filled score matrix,and finally adopts the strategy of average weighting,Predict recipe scores and recommend high score recipes to users.(3)Put forward the recipes recombination model with priority of nutrition balance and preferences.Based on the Chinese dietary intake of nutrition,combined with the user's personal information,the nutrient content table of the recipe was used as the solution space to solve the problem by using the multi-objective genetic algorithm.The recipes that meet the requirements of nutritional intake and the corresponding consumption were recommended.