Abstract: Traditional LFM didn't consider the database dynamic update which would affect the program recommendation results，so a weighted  LFM  was proposed based on the dynamic update mechanism to recommend the  personalized product service plans. Firstly，aiming at the problems of  cosine similarity neglecting individual differences，a comprehensive  similarity weighted by cosine similarity between cloud drop distance  measure and cloud was proposed to predict and fill vacancy data and reduce  data sparsity.Then，the difference of interests of new users could be  restrained and the recommendation accuracy could be improved by using the  weighted LFM recommendation service scheme. Finally，the proposed method  was updated with the method of difference averaging，and the efficiency of  different recommended algorithms was compared with mean absolute error  （MAE）. The practicality and validity of the proposed algorithm were  verified by examples.

Key words: personalized product service plan, cloud drop distance measure, cosine similarity, weighted latent factor model（LFM）

摘要： 针对传统隐语义模型（LFM）未考虑数据库动态更新从而影响方案推荐结果的问题，提出动态更新机制的加权LFM用于推荐个性化产品服务方案。针对余弦相似度 计算忽略个体差异的问题，提出采用云滴距离测度与云的余弦相似度加权后的综合相似度，预测并填充空缺数据，减少数据稀疏性；采用加权LFM推荐产品服务方案，以约束新用户兴趣差异性，提高推荐精度；采用差值平均法更新推荐结果。

关键词: 个性化产品服务方案, 云滴距离测度, 余弦相似度, 加权隐语义模型