Authors : Munir Naveed and Muath Alrammal
Abstract: In this study, a new machine learning algorithm called “RBCM” is presented to generalize the context-awareness problem for high dimensional and partially observable domains. The approach uses a non-linear regression method to build a non-deterministic probability function for generalization. A machine learning model is used to generalize the action schema given in the domain with a context. Monte-Carlo simulations are used to map high dimensional spaces to contextual spaces. A non-linear regression based value-function is applied on contextual spaces to classify the contexts into class labels. The performance of RBCM is measured using Youtube movie benchmark dataset. Youtube movie dataset has links to movies and preferences of users about the movie pairs. The movies are mapped to contexts and then contexts used to predict a user will like a movie or not. The approach is compared with the current sophisticated machine learning models. These models are decision tree (J48), bootstrapping (Ada) and Naive-Bayes. The experimental outcomes reveal that RBCM performs significantly better than its rival models.
Munir Naveed and Muath Alrammal, 2017. Reinforcement Learning Model for Classification of Youtube Movie. Journal of Engineering and Applied Sciences, 12: 8746-8750.