E-commerce Review Sentiment Analysis System Based On Machine Learning

With the continuous development of e-commerce and the increasing number of online shopping users,more and more scholars have devoted themselves to the study of recommendation algorithms for e-commerce platforms.Although some achievements have been made,there are still some remaining problems to be solved,such as data sparsity,cold startup and low recommendation accuracy.This article first analyzes the recommender system research significance and research status at home and abroad,and introduced the concept of the recommendation system and its application,and then the theoretical basis of the commonly used recommendation algorithm,algorithm process has carried on the detailed elaboration,and analyzes the advantages and disadvantages,summarizes the defects existing in current technology,but also expounds the commonly used evaluation index of the recommendation algorithm.It provides a theoretical basis for the improved algorithm in the following paragraphs.Aiming at the problem of cold start of traditional collaborative filtering recommendation algorithm,an improved recommendation algorithm based on tag weight is proposed in this paper.Firstly,establish the user-commodity-label relationship matrix,quantify the label,build a new user preference model,introduce the concept of label weight,introduce the algorithm to calculate the labe l weight of different users,and establish a new similarity calculation method based on the label weight.Experimental analysis was carried out on the different values of the regulatory factor in the similarity calculation formula and the user's neighbor set k,and finally the optimal value was obtained when the recommended results were the most accurate.In addition,the algorithm proposed in this paper is compared with other algorithms through experiments.According to the experimental results,the algorithm proposed in this chapter can get a better set of neighbors,which makes the recommended results more accurate and alleviates the cold start problem to some extent.In this paper,a singular value decomposition model based on the relationship of social friends is proposed to study the matrix decomposition algorithm of user rating matrix.The algorithm is combined with the user's social information,the preference information of friends is combined with the user's own preference information to optimize the matrix decomposition model,and the matrix is decomposed by stochastic gradient descent method.Finally,the proposed algorithm is verified and analyzed by experiments.The experimental results show that the SVD model proposed in this chapter,on the one hand,can improve the data sparsity problem to some extent,and at the same time,it is more accurate than the traditional algorithm recommendation.Finally,this paper combines the two recommendation algorithms proposed to build an ecommerce platform personalized recommendation system.According to the functional requirements of the system,the detailed design of the main functional modules was carried out,and then the modular design method was used to introduce the detailed functions and implementation of each functional module of the recommended system.The recommendation system proves the practicability of the two recommendation algorithms proposed in this paper.