Research And Application Of Recommendation Model Based On Collaborative Filtering

With the help of the Internet,data resources grow exponentially every day.In order to help users find the information they want and reduce the information overload,the recommendation algorithm arises at the historic moment.Collaborative filtering algorithm is a relatively successful application of recommendation algorithm,both in theory and practice in the research are more mature.But there are still some problems such as sparsity of data,lack of scalability,how to measure similarity among users,lack of pertinence to target users,and less application fields.This paper discusses the domestic and foreign research and application status of recommendation algorithm,and expounds the features and application conditions of the basic principle of collaborative filtering recommendation algorithms and evaluate the recommendation algorithm and each model,this paper will examine how the collaborative filtering algorithm is applied to the insurance company's insurance business,insurance business users in accordance with the purchase of insurance can be divided.into a single user insurance,two single insurance,three single insurance user user and so on A single user insurance base is huge,so as the target user.The traditional collaborative filtering algorithm is mainly applied to the field of Internet,due to the promotion of the Internet recommend areas of the implementation of the cost is very low,so there is no specialized division of high quality of the user,but recommended products for all users.Taking into account the insurance company's insurance products through marketing personnel specialized marketing,so the cost is much larger,and the promotion of speed and cycle will be longer.In order to solve this problem,this paper proposes a collaborative filtering algorithm.First of all in the index selection using statistical analysis and analysis of experts selected the risk of user purchase behavior characteristic index,using entropy method,correlation coefficient method and neural network method to determine the index weight coefficient,which makes weight calculation more scientific and reasonable,provide premise and guarantee for the subsequent modeling.Then calculate the relative possibility of the risk of the user again to buy insurance products using the ideal point method,and thus be able to sort all users according to probability again to buy insurance products,so that the insurance company marketing personnel of precision marketing,improve the marketing effect.In this paper,the target user and classifying users into loyal users,the similarity of the target user and loyal users into the collaborative filtering algorithm,through the breakdown of user groups can not only improve the success rate of the collaborative filtering algorithm is recommended,the implementation of the algorithm is improved,the better collaborative filtering algorithm application in the field of insurance,improve the company at the same time reduce the performance promotion and marketing costs.Then,the improved collaborative filtering recommendation algorithm is compared with the traditional collaborative filtering,and it is proved that the proposed algorithm is reasonable and scientific.Experiments show that the improved collaborative filtering algorithm solves the company's priority to locate high-quality user's business needs,but also to meet these high-quality users recommend the appropriate insurance products.The practical research of collaborative filtering algorithm also provides a good idea for other fields to choose the appropriate recommendation algorithm based on their own data characteristics.