基于时间与区域粒度的农资协同过滤算法

相似度计算是基于用户的协同过滤算法中的一个关键步骤，随着用户数的增加，相似度的计算空间会越来越庞大，同时在将其运用到农资领域个性化推荐时准确度较低.针对这些问题，结合农资受季节和地理位置影响强的特点对原有相似度计算方法进行改进，提出了基于时间与区域粒度的农资协同过滤算法-TA-ACF （Agricultural collaborative filtering algorithm based on both time and area）核心思想是根据已有的农资需求调研结果，建立时间与区域粒度矩阵，据此构造此时间与区域粒度内的用户评分矩阵.实验结果表明，与基于用户的协同过滤推荐算法相比，TA-ACF能够在保证时间效率的前提下，较好的提高推荐的质量.

Agricultural Collaborative Filtering Algorithm Based on Both Time and Area

Similarity calculation is a key step in the user-based collaborative filtering algorithm. As the number of users increases, the similarity computing space will become increasingly tremendous. At the same time, the accuracy is relatively low when it is applied to the agricultural personalized recommendation system. According to the feature of agricultural materials which are strongly influenced by seasons and locations, TA-ACF(Agricultural collaborative filtering algorithm based on both time and area) is proposed based on time and area size, which improves the original similarity calculation method. In this way, these above-mentioned problems could be solved. The main idea is to establish the matrix of time and size according to the existing research of the agricultural demands, and establish a rating matrix within the time and size. As the result shows, compared to the user-based collaborative filtering algorithm, TA-ACF is able to improve the quality of recommendations without losing time efficiency.