Item Selection Recommendation System Based On Collaborative Filtering Technology

The recommendation system is a large branch of the machine learning field.The design of the recommendation system refers to the design of the recommended system model or algorithm.This paper mainly studies the collaborative filtering algorithm and deep learning fusion model in the recommendation system field.The recommendation system design takes part of the information of the user and the product as input,and obtains a list of product recommendations for the current user through the design model.This paper mainly solves the problem of user cold start in the collaborative filtering recommendation system.In order to solve the above situation,we propose two collaborative filtering recommendation system models that integrate deep learning,commodity embedded collaborative filtering recommendation system model and implicit feedback collaborative filtering recommendation system model.This article mainly includes the following aspects:We propose a new method of commodity embedding,and use neural network to obtain a neural network collaborative filtering recommendation system based on implicit feedback.The recommendation system processes the implicit data and deeply explores the potential relationship between the user and the product.By improving the text embedding algorithm,the product embedding method is been obtained,and the user purchases all the products and the target goods are purchased in pairs with other products.The number of times is been mapped into a depth feature space.Compared with other related recommendation systems,the recommendation system uses two neural networks with shared parameters to obtain a target user recommended product sequence.We propose a collaborative filtering algorithm based on deep learning neural network to propose an improved scheme for the cold start problem in the recommendation system.Compared with the traditional combination method,this model not only considers the interaction information of “user and commodity” but also considers the “commodity and commodity” correlation information as the model input.The model uses the three characteristics of the user feature,the target product feature,the product set purchased by the user and the target product together as the input vector of the model.In the interaction part between the user and the commodity,the model uses the latent factor to calculate the potential vector of the user and the target commodity,inputs the feature vector of the target commodity and the product collection purchased by the user,outputs the similarity of the commodity,and finally,obtains the target through the mean pooling.The total similarity between the item and the user purchased the item.At the top of the model,the two-part calculation results are used to obtain the probability that the user is interested in the target commodity.Compared with the traditional collaborative filtering algorithm,the two models proposed in this paper prove the effectiveness of the current two models in solving the cold start problem in the recommended system task.We use the negative sampling technique to solve the complexity of the training model,which greatly improves the training efficiency of the model.