Spark-based Distance Weighted Recommendation System Research

For the traditional recommendation system for big data computing, the processing speed is slow, the recommendation accuracy is not high, and the proximity principle cannot be reflected when recommending tourist attractions. The advantages of Spark parallelized computing and memory iteration operation are used, at the same time, the user similarity is added to the nearest neighbor characteristics and the weighted coefficient of the user score multiplied by the geographical location of the scenic spot, and the weighted recommendation system model based on Spark geographic location is proposed. According to the data set provided by the Tianchi data of Alibaba cloud, the system is 2.45 times faster than the traditional recommendation system running under MapReduce, and the recommended accuracy is significantly improved.