Combining wAMAN and matrix factorization to optimize one-class collaborative filtering and its application in an emotion-aware movie recommendation system
With the development of modern science, the exponential explosion of information makes it difficult for people to find useful information from such a huge information set. As a consequence, different algorithms emerge to help people build an intelligent recommendation system. However, different algorithms have their own pros and cons. In this paper, our raw dataset is small and sparse, which contains 18 users and 275 movies. We analyze the performance of seven algorithms, but based on the performance concerning these algorithms, each algorithm does not perform equally well. We propose our new solution which can be applied in a sparse user-item matrix. An algorithm using matrix factorization by treating all missing data as negative with some weight (wAMAN) has been embedded in recommendation system. Experimental result shows that our recommendation system can find the appropriate value of the negative example in our sparse data set efficiently and possess higher accuracy comparing to the result obtained by other traditional algorithms. In addition, based on our result, we extend our design by applying it on an emotion-aware recommendation system.
ACM International Conference Proceeding Series
First Page Number
Last Page Number
Tang, Mingyue; Xie, Hang; and Tang, Tiffany Y., "Combining wAMAN and matrix factorization to optimize one-class collaborative filtering and its application in an emotion-aware movie recommendation system" (2018). Kean Publications. 1502.