To solve the problems of "new user" and "sparseness", we introduce the concept of commodity gene. Through coupling the commodity gene database, users' purchasing historical records, users' content of online browsing and the data of neighbors' behavior, we can form the module of candidated sets of customer preferences, and then use genetic algorithm which has be improved to make the selection and polymerization to the model, so that we can complete the best selection of neighbors. Finally we can get the recommended sets according to the recommended module. Experimental results show that the algorithm we suggested can improve the accuracy of the recommendation and achieve good quality of recommendation.%为解决“新用户”和“稀疏性”问题,引入商品基因的概念,通过将商品基因库、用户历史行为库、用户在线浏览内容及邻近用户行为数据耦合,形成用户偏好度候选集的兴趣模式抽取模块,然后利用改进的遗传算法优化模块进行模式选取与聚合,完成最优邻居的选择,最后经由推荐模块产生最终的推荐项目集.实验结果表明,提出的算法提高了推荐的准确度和覆盖面.
展开▼