Implementación de un sistema predictivo para grandes volúmenes de datos bajo tecnologías Big Data

Abstract

Abstract:The present project is oriented at implementing a Recommendation Web Service based on collaborative filtering for large volumes of data on Big Data platforms that allows the rapid processing of information and supports decision making. Currently on the Web, there are few automated processes to dynamically serve personalized content of products based on the preferences and interests of users and to facilitate the decision to choose one or the other product. We experimented with the MovieLens data set, the implementation of the ALS collaborative filtering algorithm of Apache Spark MLlib to make online movie recommendations using a RESTful API with Flask and, Hadoop HDFS. To evaluate the behavior of the algorithm load tests are performed to the web service obtaining recommendations of top movies, individual ratings predicted for a certain movie by adding ratings to movies by users. The results show how he is able to make recommendations online providing quick answers, but he faces the problem of cold start for new users with times of 2, 25 and 95 min.