Personalized Diet Recommendation Algorithm Based On Probabilistic Matrix Factorization With Adaboost For Diabetes

The recommendation algorithm is one of the best solutions to the problem of information overload.With the rise and the rapid development of the Internet,users who face a huge amount of data seems helpless,unconsciously already has been submerged in massive data.Due to the birth of the recommendation algorithm,the users who hover in the vastly increased data is rescued,the users' need for available information is met,the users get rid of the dilemma that they can not get the data they really need,and the utilization of information is improved.Recommendation algorithm is based on the user's interest,demand information and other attributes,which recommends the user dose not understand,the user is interested in the other information to the user.And use these attributes to associate users.After the classification of users,due to the recommendation algorithm analysis of the points of interest,so that users are closely related to each other.This personalized service,allowing users to rely on it.Recommendation algorithm has been applied in many fields,the most representation is the field of e-commerce,and recommendation algorithms in this area is developing rapidly and maturing.In the field of intelligent medicine,the recommendation algorithms have also attracted attention.Personalized diet recommendation algorithm for diabetic patients is one of the important research topics in the field of intelligent medicine.Traditional diabetic diet recommendation algorithms face more and more patient data and are unable to precisely match the relationship between the patient's body index and the recommended food.On the basis of diabetic food exchange,most existing diabetes diet recommendation algorithms mostly use association rules-based recommendation algorithm,content-based recommendation algorithm,hierarchical analysis-based recommendation algorithm,collaborative filtering recommendation algorithm,constraint-based recommendation algorithm,completed under the supervision of the attending physician and under the control of a nutritionis and so on.However,various diabetes diet recommendation algorithms described above encounter many problems in the personalized diet recommendation of diabetes,for example,ignoring the various indicators of the physical condition of diabetic patients,the method based on expert diagnosis can only understand the condition of an individual diabetic patient,the existence of inefficiency,it is wrong to think that physical indicators of people with diabetes have commonalities.Till now,there are some shortcomings in those diabetes diet recommendation algorithms,but they make some contribution to the field of intelligent medicine.However,it still has very important theoretical and practical significance to improve and innovate the research on personalized diet recommendation algorithm for diabetes.In this paper,in view of the inadequacy of the above diabetes diet recommendation algorithms,we propose a personalized diet recommendation algorithm based on the diet preference features of diabetic patients,which is based on personalized diet recommendation algorithm based on probabilistic matrix factorization with Adaboost for diabetes.The method uses probabilistic matrix factorization to associate the attributes of diabetic patients with the attributes of foods to form a matrix,extract the implicit factors that contribute to the dietary preferencesof diabetic patients and food characteristics from the correlation matrix,and then use Adaboost classifier to boostimplicit factors from a set of weakimplicit factors to a set of strong implicit factors with contributionstep by step.At the same time,we use Adaboost to train the error bound to filter out and get rid of the non-conformity weak classifiers.Weak classifiers with small classification error rate are given high weight and plays a greater role in contribution.Giving a low weight for weak classifiers with large classification error rate,they play a minor role in contribution.Compared with the above traditional diet recommendation algorithm,this algorithm can effectively balance the weights of medical indicators and individualized criteria in the field of diabetic diet recommendation,and avoid inaccuracy and incompleteness of the recommendation caused by certain factors,and improve the recommendation effect.Personalized diet recommendation algorithm based on probabilistic matrix factorization with Adaboost for diabetesconsiders a variety of extremely complex factors,which is relatively comprehensive.In the diabetic diet recommendations,it can maintain the feasibility,accuracy and interpretability.