Research On The Recommendation Algorithm For Electronic Devices Based On Collaborative Filtering

In today's rapid development of Internet technology, the amount of information we can get from the network increasingly large, at the same time we must also face the problem of information overload, difficult for users to quickly find their own valuable content from the mass of data. Recommendation algorithm application is a good solution, its use can be intelligent and active screening from large amounts of data in line with user habits data, recommended for users to meet their needs. Electrical equipment industry plays in many construction and the production of an important position in the design of electrical equipment companies face a large number of the device also has a difficult choice question, the number of projects and an increasingly large number of products, how to pick and choose the device to meet the project requirements into a serious problem. Tradition of stand-alone device management database performance limitations and storage bottlenecks in the database, has been unable to meet the current demand data to calculate the amount.Based on the collaborative filtering algorithm key technologies in-depth research, design similarity to an improved calculation method, and as a basis for the design of improved collaborative filtering algorithm, the algorithm achieved an average absolute error indexes must improve. Then combine Hadoop distributed computing system based on high-performance computing capabilities, as well as MapReduce programming framework, designed a recommendation based on the rating information for users of electrical equipment of the device prototype system designed to solve the electrical equipment, the problem of the device is recommended. The main contents are as follows:(1) In distributed computing systems and recommendation algorithm design was the basis of in-depth research and analysis above, it is designed to Hadoop distributed computing platform and the similarity of the recommendation system based on improved collaborative filtering algorithm, and We propose an improved algorithm design, using MATLAB simulation platform algorithm for the simulation based on real data sets.(2) improved collaborative filtering algorithm program conducted in-depth research, collaborative filtering algorithm is a key step in the algorithm for calculating the similarity of the proposed unified multi-dimensional quantitative indicators of Jaccard similarity algorithm(Jaccard Uniform Dimensions, JUD), and as a basis for the design of the similarity of collaborative filtering recommendation algorithm based on improved. Finally, the algorithm for the accuracy of the factors influencing the number of neighbors and other users of the relevant experiments to prove the algorithm to enhance the average absolute error and other indicators.(3) Based on MATLAB simulation platform for improving the proposed algorithm simulation experiments on real data sets, and in recent years the field of advanced algorithms were compared with the simulation results show the feasibility and reliability of the algorithm to perform well It has better recommendation results.