Movie Rating Prediction Analysis on Netflix Using Support Vector Machine Algorithm

Abstract

The rapid rise of digital streaming platforms such as Netflix has created a demand for accurate movie rating prediction systems to enhance user experience and content personalization. This study investigates the performance of the Support Vector Machine (SVM) algorithm in classifying movie ratings using historical metadata, including genre, duration, release year, and country of production. A quantitative approach is employed using RapidMiner for data mining and model evaluation. The dataset, sourced from publicly available open-access repositories, underwent preprocessing steps such as normalization, transformation of categorical attributes, and data splitting. Experimental results show that the SVM model achieved an accuracy of 83.10%. Furthermore, positive precision reached 82.98% and recall reached 100%, yielding a positive F1-score of 90.61%. However, the model exhibited limitations in detecting negative ratings, as indicated by a negative recall of only 3.98% despite achieving 100% negative precision, resulting in a low negative F1-score of 7.65%. These findings suggest that while SVM performs well in identifying positive instances, further improvement is needed for balanced classification in real-world movie rating prediction systems.