| In recent years,with the rapid development of the Internet,"finance" the Internet has become an emerging field,among them,the P2P(Peer to Peer)credit is the most typical of the Internet financial,a model for the development of the most widely,which the lender and the borrower directly via the Internet connection,with the help of a professional system platform to help both parties establish lending relationship and complete the transaction procedures.Since 2010,our country's credit P2P platform appeared a spurt of growth,due to the different credit platform operation inevitably led to a P2P credit platform launched by a lot of differences between credit products.Faced with such a wide variety of P2P credit products,investors are often difficult to in the shortest possible time to choose a most suitable for their own investment products,this problem in more professional terms is called "information overload".In order to solve this kind of problem,once appeared on the market many P2P credit use vertical search platform.Obviously this solution can only be solved when the P2P users for their own credit system of useful only when demand is very clear,in fact,in most cases,the user doesn't know what are their needs,at least not so sure.So in this case,you can often use recommendation algorithm was proposed to recommend products suitable for their credit for the user.So the recommendation algorithm is good or bad will directly affect the recommendation quality of results.At present,the recommended algorithm in the related application field in use process,there are two main problems:the first one is about the accuracy of the recommendation algorithm,in pursuit of recommended results more accurate,many researchers have been working towards this goal.Mainly responsible for the accuracy of the recommendation algorithm largely due to the users and the lack of product related data,the data matrix is extremely sparse directly influenced the accuracy of the recommendation algorithm recommended result.Main problems of the second is the cold start problem recommendation algorithm,cold start problem is very common problems in the recommendation algorithm,because in collaborative filtering recommendation algorithm,<user,project>for the grading of directly associated with the result of the recommendation algorithm,because if a project to evaluate its few users,so the project did not have a chance to be recommended to the user,this is clearly inappropriate in P2P credit system,and if the user is very few to evaluate P2P project in the credit system,the P2P credit system is probably can't find the user's interest,this will lead to no projects recommended to the user.On the above questions this paper proposes a modified using self-organizing mapping to optimize k-means algorithm to alleviate the accuracy problem of the collaborative filtering recommendation algorithm,another DecRec algorithm was proposed to alleviate the cold start problem in system filtering recommendation algorithm,in each of the proposed algorithm,the trials were carried out in this paper,and experimental results analysis,prove that the proposed algorithm in this paper can to a certain extent,ease the collaborative filtering recommendation algorithm accuracy and cold start problem. |
PDF Full Text Request