Research And Application Of Product Recommendation Algorithm In Mobile Environment

With the development of the times,online shopping has gradually become an indispensable part of people's daily life.E-commerce platforms have accumulated a large amount of historical data in their operations,and how to make full use of this data to improve users' shopping experience has become very important.Traditional recommendation algorithms cannot fully mine user behavior data,and the accuracy of predicting user purchases is not high.The user's behavioral data contains a lot of important information that can predict the products that the user may like and increase the conversion rate of the purchase.In order to improve the accuracy of predicting user purchases,this paper in the above background,uses data mining methods on Alibaba's publicly available dataset.First,this paper performs a visual analysis on the dataset,extracts the rules of user behavior data,and then train the classification model by constructing feature engineering and cleaning the feature data.In training,this paper proposes to use k-means-based clustering downsampling to adjust the positive and negative sample ratio to improve the training speed and accuracy of the classification model.Finally,the logistic regression model,random forest model,and gradient boosting decision tree model are adjusted to the best parameters and compare their performance in the test set,this paper found that the gradient boosting decision tree algorithm performs better in complex data.Based on the gradient boosting decision tree algorithm,a user-based collaborative filtering algorithm is proposed to supplement it.In order to slove the problems that users have limited time and energy in practice,and display scores of products are relatively small,a calculation using a user interest value is proposed.Finally,a hybrid approach based on sorting is proposed.When the recommendation lists of the two recommendation algorithms are merged,the recommendation list is generated by using the ranking based on the interest value.This paper proposes a hybrid method based on ranking.The recommendation list of the two recommendation algorithms is merged to generate a recommendation list based on the ranking of interest values.The comparison between experiments and traditional hybrid hybrid recommendation methods proves that the algorithm can not only bring new products to users,but also improve the accuracy of recommendations,and fully utilize the characteristics of the two recommendation algorithms.Based on the hybrid recommendation model,a personalized product recommendation system was designed,an Android shopping application was implemented,and the actual recommendation effect was tested.After summarizing the work results of this article,the related research is prospected,and the next step is planned to further improve the accuracy of recommendation and the user's shopping experience.