Abstract— Nowadays the Internet plays a significant role in our day-to-day life. An abundance of information is uploaded every second. The excess of information creates barriers to Internet users' ability to focus on their own interests. Therefore, numerous recommendation frameworks have been implemented to predict and provide suggestions to help users find their preferred items. However, it is difficult to find the most appropriate recommendation strategy(s), one that works best for the above issue. In this paper, we present a benchmarking experiment that is made by different recommendation algorithms on the MovieTweetings latest dataset and the MovieLens 1M dataset. The assessment focuses on four distinct categories of recommendation evaluation metrics in the Apache Mahout library. To make sure that we can control the benchmarking procedure efficiently and correctly, we also used the RiVaL toolkit as an evaluation tool. Our study shows that it is difficult to say which recommender algorithm provides the best recommendation and unfiltered datasets should be avoided in similar evaluations.
Index Terms— Recommender systems, benchmarking recommender systems, recommendation frameworks, MovieTweetings and MovieLens datasets.
Ruipeng Li and Luiz Fernando Capretz are with Western University in
Canada (e-mail: zack.liruipend@gmail.com, lcapretz@uwo.ca). Dr. L. F.
Capretz is on leave and is currently a visiting professor of computer science
at New York University in Abu Dhabi/UAE. (e-mail: zack.liruipend@gmail.com, lcapretz@uwo.ca).
[PDF]
Cite: Ruipeng Li and Luiz Fernando Capretz, " Assessing the Performance of Recommender Systems with MovieTweetings and MovieLens Datasets," International Journal of Innovation, Management and Technology vol. 10, no. 6, pp. 229-234, 2019.