针对传统协同过滤推荐算法不适用于情景因素，严重影响用户行为的这类场景，提出一种融合情景的推荐算法，并将该算法应用于美食推荐。首先，运用由情景属性构造向量表示情景，将情景信息作为一个重要因素添加到兴趣模型中，从而产生U-I-C兴趣模型。根据用户在不同情景下使用方式的不同，重新创建当前用户与各情景相对应的子用户，得到以情景作为标识的用户-项目评分矩阵。针对融合情景的兴趣模型易产生数据稀疏问题，设计利用改进的W-SlopeOne算法对未知评分进行填充；并通过对相似度公式进行优化，进而更加准确地找到当前用户的近邻，为用户提供更加有效的推荐服务。最后，通过实验验证该算法的有效性。%As the traditional collaborative filtering recommendation algorithm didn’ t consider the situation that context information affected the users’ behavior seriously, a recommendation algorithm with context was put forward and the algorithm was applied to food recommendation. First of all, the context was added to the traditional user-item model expressed as an attributes vector, resulting in a U-I-C interest model. Then Sub-Users were created according to different context from one user, thus obtaining a new user-item ratings matrix in a certain context. Because the data sparseness problem was easy to generate in this approach, W-SlopeOne algorithm was designed to predict unknown ratings. Based on optimized similarity formula, more effective recommen-dation service would be provided that can more accurately find the current user’s neighbor, providing users with good service. Last, experiments were done to verify the contents this paper proposed and expectation for further research was brought forward.
展开▼
