基于ALS协同过滤及频繁项挖掘的混合推荐算法

在信息技术日益发展的当下,不同类型的推荐算法在互联网行业的各领域有着广泛的应用。在目前使用较多的推荐算法中,部分侧重于基于评分的预测,也有部分是基于关联排序生成的推荐序列。本文设计一种混合推荐算法,将交替最小二乘法(ALS)计算的具体评分与Fp-growth的置信度相结合,融合两种算法的优势,从而实现推荐结果优化。基于Spark计算框架的实验表明,在选取合适的算法参数的情况下,这种改进的算法与交替最小二乘法(ALS)的原始结果相比,效率有较明显的提高,能够更准确的为用户做出个性化推荐。

Hybrid Recommendation Algorithm Based on ALS Collaborative Filtering and Frequent Item Mining

With the current development of information technology,different types of recommendation algorithms have been widely used in various fields of the Internet industry.Among the recommendation algorithms that are currently used frequently,some focus on the prediction based on score,and some are based on the recommendation sequence generated by the association order.In this paper,a hybrid recommendation algorithm is proposed,which combines the specific score of the alternating least squares(ALS)calculation with the confidence of Fp-growth,and combines the advantages of the two algorithms to optimize the recommendation results.Experiments based on the Spark calculation framework show that compared with the original results of alternating least squares(ALS),the improved algorithm can improve the efficiency more accurately when the appropriate algorithm parameters are selected,and can make personalized recommendations for users more accurately.