【摘要】 传统的协同过滤算法面对稀疏数据时计算相似度精确度偏低,导致评分预测结果计算不准确,推荐效果也随之下降.针对传统协同过滤算法的不足之处,提出了一种结合用户兴趣和改进的协同过滤推荐算法.该算法将用户评分分值差异度和用户评分倾向相似性加入到传统的协同过滤推荐算法中,同时用标准差来反映用户评分的离散性,并将离散系数作为判断用户评分对计算相似性的贡献度,与Pearson相关系数进行结合,从而消除项目自身质量属性对计算相似度带来的误差.最后将用户兴趣和改进的Pearson相关系数相结合,从而计算出更加准确的用户相似度.在真实的数据集上进行了实验验证,该算法提高了评分的预测效果,提高了推荐的精确度.更多还原

【Abstract】 The traditional collaborative filtering algorithm has lowaccuracy in calculating similarity when faced with sparse data,which leads to inaccurate calculation of the score prediction result and the recommendation effect also decreases. Aiming at the shortcomings of traditional collaborative filtering algorithm,a combining user interests with improved collaborative filtering recommendation algorithm is proposed. The algorithm adds the user score difference and user score tendency similarity to the traditional collaborative filtering recommendation algorithm,and uses the standard deviation to reflect the discreteness of the user score,and uses the discrete coefficient as the judgment user score to calculate the similarity. Contribution,combined with the Pearson correlation coefficient,to eliminate the error caused by the project’s ow n quality attributes on the calculation of similarity. Finally,user interest and improved Pearson correlation coefficients are combined to calculate more accurate user similarity. Experimental was carried out on the real data set. It showed that the algorithm improved the prediction effect of the score and improved the accuracy of the recommendation.更多还原