A Food Recommendation Algorithm Based On User Implicit Feedback

As the time for data concentration is coming,people's increasing mobility led them gradually into the ocean of information.How to provide users that have different locations and different interests background with personalized recommendation service is a challenging research subject.Most of existing research focus on electronic commerce,recommendation for books and audio,But a few research achievements on personalized recommendation for food industry.Meanwhile,most of existing research based on the User-Item dual structure,and there is no location information,Single commodity classification is difficult to meet the requirements of class such as food recommendation applications.After analysis the traditional personalized recommendation technology and compare to the food industry characteristics,this topic design a kind of personalized food recommendation algorithm based on implicit user feedback.First of all,in this thesis,this article review the current personalized recommendation algorithm development status at home and abroad,and comparative analysis the mainstream recommendation algorithm and user model.Secondly,collect user behavior by record user interest explicit and implicit,combining location information and using collaborative filtering and interest clustering,ect.Commodity forest hierarchy design a multidimensiona space model(user-item-shop),and build the model separate for shop with spatial attribute and item with non-spatial attribute,and integration of the user's implicit data.Compared with the traditional recommendation model,through by collect user's implicit data and improvement of model.it has greatly improved the efficiency of recommendation.Then,an personalized recommendation algorithm aimed at the characteristics of food was designed based on the improved model,which mainly included user similarity and item similarity recommendation algorithms and rating prediction algorithm,according to ameliorate model and algorithm,the recommend accuracy has improved and realized seamless link which from the personalized items recommendation to the personalized merchants recommendation.Finally,the thesis verifies the performance of the improved model and algorithm through by simulation experiments and prototype system,which included the compared about the similarity of user and the similarity of item and recommendation of shop,etc.Experimental results show that using the personalized recommendation algorithm of this model can effectively enhance the recommendation accuracy and efficiency.Meanwhile,the algorithm has a good reference value on industries applications similar to food industry.