Research On The Recommendation Model Of Customer Purchase Problem For E-commerce

In recent years,online shopping has become more and more important in people's daily lives.The e-commerce services are diversified.When customers purchase products on the ecommerce platform,faced with huge information,there will be problems such as decreased shopping efficiency and difficult shopping decisions.Recommendation system can help customers reduce shopping time and meet their individual needs.This article compares the main recommended models of existing e-commerce platforms.The content-based recommendation system is intuitive and interpretable,but it is vulnerable to new customers and new products.The recommendation system based on association rules is easy to find new preferences,but rule extraction is more difficult;recommendation system based on collaborative filtering The recommendation effect is good,and the recommendation effect is continuously improved over time,but it is vulnerable to data sparsity.Based on an overview and comparison of the above three models,it is found that the current recommendation system used by the e-commerce platform basically only considers the customer's historical behavior data,and there is less research on the online comments filled by customers.Currently online reviews have become an important basis for customer decision-making.Therefore,online reviews also have high application value for e-commerce recommendation models.The commentary contains the customer's personal emotions and detailed description information on the characteristics of the products.These commentary data are very helpful for explaining the customer's purchase behavior of the product.This article combines the characteristics of online comment data,taking large-scale e-commerce platform clothing products as an example,proposes a recommendation model based on online reviews.The model first carries out text mining of online comments,obtains online reviews of clothing,shoes,and hats on large e-commerce platforms through web crawlers,segmenting text,extracts feature words using TF-IDF,and assigns emotional words based on emotional vocabulary,combined with the analytic hierarchy process for weighting,establish a final scoring system.The model is based on the customercommodity rating table,combined with commodity-based collaborative filtering to calculate the similarity of the product,and finally achieves the recommendation by predicting the customer's score on the product.This article carries on the off-line experiment to the model through the real data set of a company's website,and verifies the feasibility of the model.