Research And Implementation Of Recommendation Algorithm Based On Apache Mahout

With the rapid development of Internet,the growing number of web pages access tothe Internet. The traditional search algorithm can only be presented to all users the sameresult, can’t provide the appropriate information for different users, arising the problemof "information overload". Therefore, the personalized recommendation technologycame into being.Collaborative filtering recommendation algorithm is the most widely usedrecommendation algorithm in the current recommendation system.As the e-commerceexpanding the scale, collaborative filtering algorithm also face a number of challenges.Such as cold start problem and data sparsely problem.This essay conducts in-depthstudy and research of collaborative filtering algorithm, and proposed combinationalgorithm and improved algorithm for collaborative filtering algorithm, achieved thedesired results.The main study of this essayis as follows:1. It has a detailed understanding of the status of the recommendation system andrecommendation algorithm. Focus on collaborative filtering recommendation algorithmand Apache Mahout recommendation algorithm knowledge. It has an introduction forcurrent mainstream recommendation system and recommendation algorithm, and theadvantages and disadvantages of the various recommendation algorithms.2. Carried out a detailed analysis of the collaborative filtering recommendationalgorithm.The algorithm mainly includes two categories, they are user-basedcollaborative filtering recommendation algorithm(User-Based CF) and item-basedcollaborative filtering recommendation algorithm(Item-Based CF).Also focus onresearch the widely used Slope One recommendation algorithm, andintroduce theprinciples and steps of the three algorithm detailed analysis.3. Proposed a combination recommendation algorithm.The algorithm is based onthe item-based collaborative filtering recommendation algorithm and user-basedcollaborative filtering recommendation algorithm combination.Take full advantage ofthe user-item rating data set contains the users and projects torecommend.4. Use Apache Mahout open-source framework, MovieLens data sets and MAEevaluation criteria, having simulation experimentthe traditional item-based collaborative filtering algorithm,user-based collaborative filtering algorithm and SlopeOnecollaborative filtering algorithm. Compare the recommendation resultof thethreemethod of calculation of the similarity. And it has a simulation experiment for thecombination algorithm. Contrast the traditional collaborative filtering algorithm,improved collaborative filtering algorithm and combinatorial algorithms experimentaleffect. Analyzethe experimental results.