基于隐式数据和Apriori的协同过滤推荐算法

Collaborative filtering recommendation algorithm based on implicit data and Apriori algorithm

针对传统协同过滤推荐算法对目标客户进行个性化推荐时,因用户评价数据和物品属性等显式数据稀疏,造成推荐商品的准确率和质量相对较差的问题,本文基于隐式数据和Apriori算法对协同过滤推荐算法做出改进。首先,算法基于隐式数据中用户对商品的行为和用户对商品的评价,建立用户对商品的评分偏好模型,用以构建原始评分数据;其次,利用Apriori算法找出用户行为数据集中商品的强关联规则,利用输出的关联规则对原始评分数据进行降维,并进行相似度计算,确定用户之间的相似性,根据计算结果来确定目标用户的近邻集合;最后,算法通过度量后的最近邻居来计算目标用户对特定商品的预测评分。从数据集中分别采取70000条数据和30000条数据进行算法测试,测试结果表明改进后的推荐算法与基于用户的协同过滤算法相比准确率和召回率分别提高了1.56%和0.23%;和基于项目的推荐算法相比准确率和召回率分别提高了4.39%和0.92%,证明基于隐式数据和Apriori算法改进的协同过滤算法,在缓解数据稀疏的同时,能提高推荐的准确率。

In order to improve the traditional collaborative filtering recommendation algorithm in the personalized recommendation of target customers,the accuracy and quality of recommended products are relatively poor due to sparse explicit data such as user evaluation data and item attributes.This paper improves the collaborative filtering recommendation algorithm based on implicit data and Apriori algorithm.First,the algorithm builds the user’s scoring preference model for the product based on the user’s behavior on the product and the user’s evaluation of the product in the implicit data to construct the original score data.Second,the algorithm uses Apriori to find the strong association rules of the products in the user behavior data set.The output association rules are used to reduce the dimensionality of the original score data,and the similarity calculation is performed to determine the similarity between users,and the neighbor set of the target user is determined according to the calculation result.Finally,the algorithm uses the measured nearest neighbors to calculate the target user’s prediction score for a specific product.The experiment takes 70,000 pieces of data and 30,000 pieces of data from the data set for algorithm testing.The test results show that the improved recommendation algorithm has increased accuracy and recall rate by 1.56%and 0.23%compared with the user-based collaborative filtering algorithm,and Compared with the item-based recommendation algorithm,the accuracy and recall rate are increased by 4.39%and 0.92%,respectively.Experiments prove that the collaborative filtering algorithm based on implicit data and Apriori algorithm improves the accuracy of recommendation while alleviating data sparseness.