Research And Implementation Of Apparel System Based On Spark Machine Learning

With the rapid development of Internet technology,e-commerce sites and mobile Internet,the number of goods and usersine-commerce sitesappeared a rapid growth trend.When e-commerce sites face growing users and goods,how to dig out the useful information from the massive data,how to improve customer loyalty and product sales is facing a serious problem of electronic commerce website.Machine learning and recommendation system can solve this problem effectively.In the recommended system,collaborative filtering recommendation technology is widely used as the most typical recommendation technology.It can be personalized recommendation according to the user's preferences and interests.However,there are also some problems in collaborative filtering recommendation algorithm,such as scalability,cold start,sparseness of data and so on.The scalability problem is due to the increase in the amount of data,the cost of the recommended system gradually increases,and t he time and space complexity of the system will be more and more.The main reason of data sparsity is that the user's behavior data has a large sparsity,which leads to the low accuracy of collaborative filtering algorithm.In order to solve the above proble ms,this thesis proposes a hybrid recommendation algorithm based on Spark machine learning,The hybrid recommendation algorithm can effectively improve the accuracy of the recommendation effect.and finally the hybrid recommendation algorithm is applied in the apparel electricity supplier website.In the traditional recommendation algorithm,the stand-alone machine in dealing with massive user behavior and commodity data need to spend a lot of time,the parallelization of the stand-alone algorithm can solve this problemIn this thesis,the hybrid recommendation algorithm is applied to the cluster based on the Spark platform,and the parallel recommendation and optimization of the hybrid recommendation algorithm in the cluster are realized.Not only solve the problem of computational efficiency,but also solve the scalability problem.Then,this thesis designs a clothing electricity supplier system,which includes user,businesses and system administrator,the user can purchase clothing,management orders,view shopping cart,the purchase of clothing scores and other functions;Businesses can add goods,manage the library and other functions Then we apply the hybrid recommendation algorithm in this thesis to the clothing electricity supplier recommendation system,reco mmend cloths the user might like to the user according to the user's historical behavior data.Experiments show that the proposed hybrid recommendation algorithm based on machine learning can effectively improve the accuracy of recommendation results.In particular,the algorithm is based on the Spark platform,which is more suitable for a large number of iterative computation process and processing massive data,the algorithm has more advantageous in running speed,and has better scalability.