Reason On Recipe Recommendation For Fusion Knowledge Graph And Collaborative Filtering

Three meals a day occupy an important position in people's lives,but people are often troubled by what to eat every day.The emergence of the Internet has brought a lot of convenience to all aspects of people's lives.It has become the norm for people to obtain information from the Internet,including recipe information.But getting what you really need from hundreds of thousands of distributed messages is a waste of time and effort.This era is an efficient era,in this information explosion,the recommendation system came into being.The recommendation system can predict the user's preferences according to the user's historical behavior and recommend to the user,which can save the user's query time and greatly improve the user's efficiency.Although the application of recommendation system involves a wide range of fields,but after investigation and analysis,there is less research in the field of recipe recommendation,and the traditional recipe recommendation system does not combine the user's preference information.Recipes recommended to users are sometimes not loved by users,so this.Collaborative filtering algorithm is the most widely used recommendation algorithm.In this paper,knowledge map connotation knowledge is added to the collaborative filtering algorithm to improve the recommendation effect.By integrating the triple information in the knowledge map,the similarity degree of the two vectors is calculated by using Euclidean distance.The research work of this paper mainly includes the following contents:(1)The Construction of the Recipe Knowledge GraphUse python's scarpy framework to climb data from three recipes health sites(Meishijie,Beut Fruit Food,39 Health Net),extract the entity data and relational data from the recipe triples,construct the corresponding recipe knowledge graph,and store the triplet information in the neo4 j graph database.(2)Improved the Trans R translation modelAdding dynamic translation to the translation model Trans R to define the embedding range of triples as any one vector in the plane range,the idea can well distinguish some similar entities or similar relations and avoid the mistakes of the traditional translation model in representation learning.(3)Recommended model for fusion knowledge graph and collaborative filtering DTransR-CF.Integrate the improved DTransR model with collaborative filtering methods into DTransR-CF.The model uses both internal relations and external information to improve the recommendation effect by adding relevant background knowledge of KG to the CF algorithm.The effectiveness of the model is analyzed by comparing the experimental results with other models.