Design And Implementation Of A Recommender System Framework Based On Collaborative Filtering Algorithm

Currently, one user can gain a lot of convenience by the e-commerce system. Lots of thee-commerce system introduce recommendation engine as the role of a salesman. Due to thewidely used of recommendation engine, the research of collaborative filtering algorithm isflourish. A suitable collaborative filtering algorithm can bring a lot of benefit. But when weface to all kinds of collaborative filtering research, most of them emphasize the algorithm butoverlook the practical use. Slope One algorithm has very good effects in use. However, itcan’t get rid of the sparse data, cold start problem by only using Slope One algorithm. Thisresearch will not limit to the algorithm and will introduce the recommender systemframework.This research has the following aspects:(1) Research the e-commerce system. Explain the significance of the recommendersystem by a real world e-commerce system and the analysis of the algorithm and the realitysystem. Describe the remaining problems.(2) Analyse the Slope One algorithm and the critical technology like Lucene, NoSQL.And finally put forward a kind of recommended way, this way can satisfy the users’ favor andcan expand the business.And discuss and improve the practical problems to make the systemin full cry.(3) To an e-commerce system, this research shows the detail requirements. This researchwill put forward the design goal of the system frame and focuses on the algorithm layer (improvement ideas and concrete realization), but not only the algorithm. This research alsodesigns the other system modules detailly, like recommended results show layer, to make thesystem scalable by using the design patten. The main purpose is to get better effects withsimple methods.(4) By using a lot of experimental data, this research will make a conclusion whichtechnology is better to fit the system.(5) Design experiments and select MovieLens data set. On the basis of classified datasets this research gets the better similarity algorithm and then proves the new algorithm canhave better effect by the test.And finally, the research points out the prospect of the recommendation system and theframework.