The Research On Recommendation System For E-commerce Based On Hadoop

E-commerce makes our life become convenient,at the same time,the E-commerce has produced a large amount of data, It has become an important issue that how to help users find valuable content quickly and efficiently in the vast amounts of data.At present, in the face of a huge number of information,the information classification website and the search engine have effectively solved the problems of search problem in the huge amounts of data, but in order to get the needed information,both of which need users to search by keyword or other relevant information. By contrast, The recommendation system is more active and intelligent, So it has played a more and more important role in e-commerce sites. It can quite fast search in huge amounts of data and don’t need users to input keywords and other tip information, recommend actively useful goods to customers, Its intelligence not only provide customers with convenient and provide with a lot of help for businesses.There a lot of arithmetics supporting the normal running of recommendation system,The collaborative filtering recommendation algorithm is one of the most important algorithm. However, As the number of users and goods in e-commerce increase rapidly, collaborative filtering recommendation algorithm is also facing new challenges, Such as data problems about data sparseness and scalability, and so on.To solve these problems, this topic conducts a comprehensive in-depth research on collaborative filtering recommendation algorithm, expounde to solve the data sparseness problem by combining recommendation algorithm, and then complete the recommended job. At the same time, considering the limited by single machine performance,It will have serious impact on the accuracy and efficiency of the recommended.result,when needs to handle huge amounts of data. So, deploying the collaborative filtering recommendation algorithm on Hadoop, and these data are adopt the method of distributed processing, improving the efficiency of the algorithm, solving the scalability problem of the algorithm, ultimately achieving the goal of increasing commodity sales.This paper main research work is as follows:1) Study and analysis the commonly used several kinds of recommendation algorithm, having a comprehensive understanding of the advantages and disadvantages of each algorithm. The research is mainly focused on the collaborative filtering recommendation algorithm.2) Using a combination of recommendation algorithm to implement the data filling and result is recommended. The combination algorithm is to combine the K-Means clustering algorithm, the Slope One weighted algorithm and collaborative filtering (CF) algorithm. The K-Means clustering algorithm and the Slope One weighted algorithm can solve the problem of data sparseness, The collaborative filtering (CF) algorithm is used to implement the final recommendations on the basis of the relatively complete data.3) Improving the collaborative filtering recommendation algorithm, and enable it to adapt to the graphs programming model MapReduce, and then complete the data distributed processing, to solve the scalability problems of the algorithm in this way.4) Evaluating a single recommendation algorithm and the combination recommendation algorithm. Using MovieLens data, evaluating these algorithm on the precision rate,the recall rate and reaction time through the experiment, and analyze the experimental results.