摘要
通过提高挖掘效率、增强算法扩展性,解决传统的推荐算法在旅游景点推荐方面响应时间长、推荐效率低,无法适应大数据挖掘需求的问题。对现有协同过滤推荐算法进行深入分析,选取适用于旅游景点推荐的Slope One算法和Itembased算法。将这两种算法高效结合,并基于MapReduce编程在Hadoop云平台上实现算法并行化,通过采集"旅评网"的真实旅游景点评分数据验证算法的有效性。通过测试真实的旅游景点评分数据,表明算法不仅提高了推荐的准确度,而且比传统的协同过滤算法具有更高的运行速度。实验结果较好地说明了该算法具有更高的挖掘性能和可扩展性,能够更好地适应旅游景点数据量大、数据矩阵稀疏的特性,满足旅游景点推荐高命中率和个性化的要求。
The problems that the traditional tourist recommendation algorithm has long response and low efficiency in tourist recommenda- tion, and have been unable to meet the mining needs of large amount of data can be solved by improving mining efficiency and enhancing scalability, It conducts a deep analysis for existing collaborative filtering algorithms in this paper and selects the Slope One algorithm and the Item-based algorithm suitable for tourist attractions recommendation to efficiently combine. Then the new algorithm is paralleled based on Hadoop framework,and the algorithm＇ s validity is proved by the data collected from ＂www. ilvping, corn＇. The faster speed and higher accuracy of the recommendation algorithm have been proved by the data collected from ＂www. ilvping, corn＂. Experimental results show that the new algorithm has a better performance and scalability, which can be better to solve the problems of big data and sparse matrix, and meet the requirement of high percentage shot and personalization.
出处
《计算机技术与发展》
2016年第3期47-52,共6页
Computer Technology and Development
基金
国家自然科学基金资助项目(71473114)
