Research On Recommendation Algorithms For Individual And Bundled Products

With the continuous development of Internet technology and mobile payment,online shopping has become essential part of people's everyday life.Recommender systems are an import role in helping users filter out their favorite products in online shopping.Personalized product recommendation is based on the user's historical behavior to recommend a list of products that meet their preferences.According to whether there is a relationship in the recommended product list,this article divides product recommendations into independent product recommendation and combined product recommendation,and conducts two questions the study.First of all,in view of the data sparseness and cold start problems in personalized product recommendation,this article formulates a variety of recommendation strategies to adapt to different types of users,the concept of dating user experience,and defines the gradual number of implicit feedback behaviors of users,and proposes An independent product recommendation model that uses user experience to balance multiple recommendation strategies,and considers that the balance coefficient is too large to cause the imbalance of multiple recommendation strategies,and the model is re-adjusted to optimize the model to improve the accuracy and variety of independent product recommendation.Furthermore,independent product recommendation does not take into account the relationship between recommended product lists,resulting in too single recommendation results,and combined product recommendations are released here to achieve diversification of recommended products.Aiming at the problem of sparse combined product data sets in existing combined product recommendation research,this paper is based on the crowdsourcing technology,through task design and quality control,explores the relationship between products and establishes a reliable bundle product data set.In order to enhance the performance of combined product recommendation,this paper proposes a bundle product recommendation model based on variational neural network collaboration filtering.The variational autoencoders model extracts the characteristics of the bundle product and predicts the score of the combined product based on the neural network collaborative filtering model.Finally,experiments were carried out on the two real-world data sets and the combined product data set of the merged product.The multi-strategy independent product recommendation model fused with user experience and the combined product recommendation model based on variational neural collaborative filtering were tested to verify the performance of the model proposed in this paper.