常昊 杨盛泉



摘要: 大数据时代的到来,生活种每天都产生着大量的数据,对这些数据进行分析并推荐商品在电商领域就会显得尤为的重要。协同过滤算法是目前发展比较成熟的,在各领域都取得了非常好的效果。但传统的协同过滤算法在计算相似度和预测评分时太过粗糙,并且效率很低。我们将协同过滤算法与决策树算法相结合,并对算法进行改进,创建一种协同过滤决策树算法来对商品进行推荐,并将新的协同过滤决策树算法运行在Hadoop平台上。Hadoop是一种分布式的处理大规模数据的云计算平台。实验证明,改进后算法使得推荐的准确率有了显著的提升。
Abstract: With the advent of the era of big data, daily life is generating a large amount of data. Analyzing and recommending these data will be particularly important in the field of e-commerce. Collaborative filtering algorithms are currently relatively mature and have achieved very good results in various fields. However, the traditional collaborative filtering algorithm is too rough and inefficient in calculating similarity and prediction score. We combine the collaborative filtering algorithm with the decision tree algorithm and improve the algorithm to create a collaborative filtering decision tree algorithm to recommend products and run the new collaborative filtering decision tree algorithm on the Hadoop platform. Hadoop is a distributed cloud computing platform for processing large-scale data. Experiments have shown that the improved algorithm has significantly improved the accuracy of recommendation.
关键词:大数据;推荐算法;协同过滤;Hadoop
Key words: big data;recommendation algorithm;collaborative filtering;Hadoop
中图分类号:TP311.5 文献标识码:A 文章编号:1006-4311(2020)09-0127-03
0 引言
随着科技的发展,互联网技术也随之迅速发展和普及,这就使得网络上的数据每天高速增长,丰富用户生活消费的同时,大量信息资源膨胀迅速的问题也随之出现,“信息过载”也是目前互联网正面对的难题。“信息过载”指的是用户难以从海量数据中准确并且迅速地定位到自己所需要的信息。推荐系统的出现大大地缓解了“信息过载”这一难题。推荐系统根据用户的历史行为,预测并且推荐使用户满意的物品,实现个性化的服务。推荐系统是自动智能的为用户推荐物品,而且还会根据用户行为的变化,动态的调整推荐的物品类型,真正意义上避免了“信息过载”的困扰。
推荐算法是推荐系统……