Design And Implementation Of Personalized Recommendation System

With the growing maturity of Internet technology, People do some activities on the Internet, such as shopping, listening, and looking some videos. As a result, more and more data is generated. As the amount of information increases, it becomes increasingly difficult for people to find the information they are interested in. Traditional approach is entering a number of key words, and then using the Search Engines such as Baidu and Google to find the useful information. However, if people can’t provide a very effective keyword, which leads to the search engine can’t quickly find useful information. It is the existence of the search engine ‘s these shortcomings. In this way, there is more and more dark information on the Internet. Therefore, a new technology- the recommended system has been produced for using the dark information effectively.Collaborative filtering recommendation system is the first recommendation system, but also the most classic and most widely used recommendation system. Collaborative filtering recommendation system can not only discover the potential of users, but they did not find the interests, but also can recommend art, music and movies and other difficult to analyze the content of the product. But at the same time, he has some difficult problems to solve, For example, how to recommend to new users, how to recommend new products; how to explain the results of the recommendation; whether the classification of the recommended products is accurate; whether the accuracy of the product prediction score, etc. Content based recommendation system can be regarded as the continuation and development of collaborative filtering recommendation system. It can not only solve the new users and new products’ problem which the collaborative filtering recommendation system cannot solve, but also better explained on the recommendation results. However, it also has the problem of classification accuracy and prediction accuracy. In recent years, the network structure based recommendation algorithm has been developed rapidly, the most representative of that is the material diffusion algorithm. Material diffusion system has better effect of the recommendation, classification and prediction are more accurate, it is relatively simple to achieve. However, it does not take into account the impact of user ratings on the recommended results.The main purpose of this paper is to improve some parameter into the initial resource matrix and transfer resources matrix which are the important two portion of the recommended system that based on the network structure. It also consider the user’s score, and then add the regulator into the algorithm. And then test the algorithm use the data sets that is generated in the Movie Lens website. It also compare the analysis of the experimental results during the user-based Collaborative filtering recommendation algorithm, the item-based collaborative filtering recommendation algorithm and the content-based recommendation algorithm and the standard material diffusion algorithm and the improved algorithm in this paper. The final results show that both the classification accuracy and prediction accuracy of the algorithm are better than other existing recommendation algorithm. In addition, this paper uses a very wide range of tools in the field of statistics and analysis of the data analysis tool R to mining and analysis the data. R is used for statistical analysis, drawing the language and operating environment. It is a good tool for statistical computing and statistical mapping.