Research On Taobao Commodity Personalized Recommendation Algorithm

The advent of the era of big data has led to an increase in information beyond expectations.The emergence and continuous development of the recommendation system has effectively alleviated this problem.However,people think that the improvement of the quality of life and the need for personalization are highlighted.How to mainly study the content of different customers' personalities.In this paper,we will study the personalized recommendation algorithm for this problem and propose two personalized recommendation models for different data sets.as follows:When recommending a compressed small sample behavior data set in the Taobao platform,there are often problems with data set imbalance and sparse samples.A new model is proposed for the classification problem of unbalanced data sets.The model first uses the TSMOTE algorithm to cyclically synthesize the small samples at the decision boundary in the unbalanced data set,and then the newly synthesized small sample samples are outside the decision boundary Synthesized sets of less-class samples coexist to perform oversampling to increase the importance of minority samples;secondly,for the phenomenon of hyperplane migration during training due to data imbalance in the SVM classification process,the DEC algorithm is used as a classification The classification algorithm of the filter and the use of objective standard deviation to select the marker coefficients improve the performance of the classification algorithm.In order to reduce the sparseness of the data set,this paper uses KNN and improved SVM to make predictions.The model first obtains two types of "like" and"dislike" through two classifications.Then through KNN and SVM hybrid collaborative filtering classification model.Perform SVM multi-class score prediction,and finally make Top-N recommendation.Due to the large amount of user behavior data on Taobao,it is difficult to obtain a good recommendation effect using a single recommendation method.Therefore,here we propose a model that integrates UserCF,ItemCF and Xgboost algorithms of time series for personalized recommendation.First,a coarse-grained recall of the data set is performed by UserCF fused with a time constant.Then use stepwise and feature design and extraction of new samples for Xgboost training and CV adjustment to obtain a trained Xgboost prediction model.Finally,make the final list recommendation.According to the characteristics of SVM suitable for processing small data sets and Xgboost integrated learning suitable for processing large data sets,and using the optimization methods proposed in this paper to test the data sets of different scenarios of Taobao products,the accuracy has been improved.The experimental results of the final classification recommendation model prove that the classification model in this paper can provide customers with decision support through accurate classification;the improved Xgboost model has improved accuracy,recall and F1 values,mean square error MSE,root mean square The comparison of the error RMSE and the average absolute error MAE value also confirms that the recommended model here has higher accuracy.