基于综合相似度与任务紧要度的个性化任务推荐

Personalized Task Recommendation Based on Integrated Similarity and Item Significance

研究了如何将协同过滤推荐应用于IT项目外包平台,实现个性化任务推荐,提出了1种融合用户Profile文本相似度、任务选择相似度及任务紧要度的协同推荐方法.该方法将用户对任务的选择行为转换为用户-任务类选择矩阵,并以此计算用户间的选择相似性;用户profile文本相似性用于平衡用户选择相似性并形成用户综合相似性,算法中任务紧要度用于度量任务的时限性与经济性,设置合适的阈值来构建待推荐任务集.在真实数据集上的实验结果表明,提出的个性化推荐方法具有较高的推荐准确度,并在一定程度上缓解冷启动与数据稀疏性问题.

The area of collaborative filtering （CF） applied in IT project outsourcing is studied and the researches on personalized task recommendation are carried out. Based on this, a personalized task recommendation method is presented combined with Profile text similarity, task selection similarity and integrated similarity. This method transforms the users＇ selection behavior into user items-class selection matrix, and it is used to compute the selection similarity among users. User＇ s profile text similarity is also considered to balance the selection similarity, while integrated similarity is applied to measure the timeliness and values and build waiting recommendation item set by choosing the proper threshold. The experimental results indicate that the proposed method is effective and it could be used to alleviate the data sparseness and cold start problems.