A Novel Approach Towards Context Sensitive Recommendations Based on Machine Learning Methodology

Contexts and aspects have been distinguished as the significant factors in fabricating recommender systems. Most recommender systems aim at utilizing either non-contextual preferences or contextual preferences distinctly, while very few endeavors have been made to identify the significance of both. Hence an attempt has been made to study the influence of both, users' context dependent and context independent preferences in the single recommender system. In this case, accuracy has always been a challenge. Therefore, there exists a need for such a classification technique which can be commonly applied to both types of preferences that helps enhance the accuracy of the retrieved results. For this purpose, use of a standard Machine learning technique well known as Support Vector Machines was proposed. The idea behind using Support vector machines is to split the data in an optimal way and classify the data precisely to aid prediction purpose. For generating recommendations, these context-dependent preferences are further combined with users' context-independent preferences. Finally this technique is applied on a real-life dataset to demonstrate that our method is proficient in dealing with contextual preferences of users and well classify them to achieve better recommendation accuracy than the relative works.