Enhanced Movie Recommendation System Using Temporal Gradient Entropy

Recommendation systems play a crucial role in many industries such as e-commerce and entertainment. Since consumers in today's world are confronted with a lot of choices, recommendation systems help in filtering choices and will improve user experience. This increases the users' satisfaction and encourages them to continue to engage with the product or service. In the current days, businesses also can take advantage from recommender systems because such systems allow the business and revenue streams to grow by improving user retention and engagement. Additionally, businesses can gather useful information and trends in the consumption of content which can then be leveraged by business intelligence for further analysis and other business opportunities. Thus, this research focuses on yet another novel approach on building recommendation algorithms which can perform more satisfactory recommendations to users by using the concept of "Temporal Gradient Entropy". This approach tries to capture the temporal dynamics and trends in user preferences and measures the changes in the user's interactions over time. The dataset used in this research to demonstrate the advantages of Temporal Gradient Entropy is the Netflix Prize dataset. This dataset is released by Netflix officially and consists of over 480K users and 17K movies with their respective movie ratings along with the date when the rating was given. The dataset is cleaned and processed by the big data processing framework Hadoop and MapReduce, after which the cleaned data was used to calculate the Temporal Gradient Entropy. We highlight the approach and showcase the results of calculating the Temporal Gradient Entropy and using this metric together with modern Machine Learning models such as Singular Value Decomposition (SVD) to perform satisfactory recommendations.