The Problem About Sparse Data of Recommended System.
Introduction of Problem StatementThe problem about sparse data of recommended system.Collaborative filtering technology can analyze customer base to form a common consumer tastes recommended. Data scarcity problem is a major challenge faced by collaborative filtering technology. ROCK clustering algorithm utilizing a recommendation system model based on collaborative filtering technology, the model can effectively solve Collaborative Filtering data scarcity problem.As electronic commerce is rapidly developing. E-commerce system information “overload” phenomenon more and more serious, face commodity information “sea”, the consumer is difficult to quickly and efficiently pick out what he needed goods. On the basis of accurate identification of customers on consumer preferences. E-commerce recommendation system can provide product information and advice to clients. Analog sales staff to help customers complete the purchase process, allowing customers to avoid trouble information “overload” brought . Recommended is to determine the accuracy of the results of the key factors to success recommendation system, if the system is recommended to customers for goods that customers do not need, then the customer will lose confidence in the system of recommendation. The recommended information as garbage information. The recommendation is based collaborative recommendation system recommendations and other means correlation between the customer depending on the target customer, when the system found one or a group of customers with similar consumption preferences target customers, the system can be predicted based on the users consumer behavior target users of consumer behavior. However, on an e-commerce site, the number of customers and the number of goods are enormous. In this case. Accurately target customers looking for a set of consumer preferences similar customer base is a very difficult problem.
Collaborative filtering problem is to predict the extent of a certain body like he has not been the object of evaluation, forecasting is based on a series of objects of historical evaluation of past record user groups.User evaluation of the object can be explicit or may be implicit. Explicit evaluation usually Ratings customers in the form of the products value. If the count value is very high. It indicates that the user likes the product, on the contrary that the user does not like the product. If you want help recommendation system. First, he needs to submit evaluation information system for some products. Implicit evaluation is usually derived from data resources towel out. For example, a user browsing time analysis of each page, the sites log file analysis, or analyzing the users purchase history. By analyzing these implicit preference information. This information could eventually be mapped to an explicit evaluation information. Whether explicit or implicit evaluation information evaluation information, the final evaluation can be mapped to a recording sheet.