Design And Implementation Of Peer-To-Peer Lending Products Recommendation System Based On Hybird Algorithms

In recent years, with the advent of the Internet Finance, In China, the P2P(Peer-To-Peer) lending platform growth has explode. Because of the different business model in each platform led to the launch of the financial products is uneven,which is filled with all kinds of financial products. For investors who faced with so many P2P lending platforms are hard to quickly find the lending products they need, and because of the current lack of strict regulation of the P2P lending industry, the results can lead to a large number of potential users and excellent lending products just miss the person or opportunity. In view of this,the market appeared some vertical search platforms focused on helping the user finds P2P lending products, but these search platforms are hard to make more efficient when the user is not clear their own needs. This recommendation system solves the information overload problem of P2P lending products and makes up for the search platforms shortcomings, it eventually makes investors and P2P platforms to achieve win-win results.First, this thesis introduces the current development status of P2P lending platforms and recommendation system,and research on the item-based collaborative filtering and content-based algorithm. Second, designed a hybrid recommendation system architecture and recommendation process base on these two kinds of algorithm, and mainly introduces the implementation principle of the system module. Finally, Evaluation for the system, summarizes the shortcomings of the system and then lists the future work. The main contents of this thesis includes:(1) Describe the system needs analysis in detail, design of the overall system architecture and hybrid recommendation algorithm, and the system divided into four modules:the spider module, the offline calculation module, the online calculation module and the data storage module.(2) The spider module adopts distributed architecture design, it can extract data from multiple P2P lending platform and support for the management of each spider agent, remote monitoring.(3) The offline calculation module uses ItemCF algorithm and Kmeans algorithm in Mahout run on the Hadoop platform, and the offline calculation results are saved to the HBase.(4) The online calculation module run on the Storm platform, and it recommends the lending products for the user according to the different recommendations.(5) Test and evaluation for this recommendation system.Base on the above discussion, this thesis builds up a P2P lending products recommendation system base on hybrid algorithm. Experiments show that by using offline calculation and online calculation of data processing capabilities, and the advantage of the hybrid recommendation algorithm effectively solves the problem of information overload, and to provide more personalized P2P lending products for investors.