A user-based video recommendation approach using CAC filtering, PCA with LDOS-CoMoDa

Abstract

Context-aware collaborative filtering is an efficient method for tailoring recommendations to the individual contextual settings of users, with the objective of improving prediction accuracy for a Context-aware recommender system. This research proposes a video recommendation model based on Context Aware-Collaborative Filtering with Principal Component Analysis (PCA). Traditional recommendation system models use user preferences and make recommendations using either content-based or collaborative filtering methodologies. These filtering methodologies suffer from data sparsity problems. To overcome this problem, instead of using conventional methods, this work uses neighbourhood-based collaborative filtering. The LDOS-CoMoDa dataset is used to evaluate performance. In this work, the PCA approach for recommender systems is developed to calculate efficient similarity across various users and based on the attributes for the video recommendations system. The primary objective is to enhance the performances of the context-aware recommendation model using a collaborative filtering method with PCA. In addition to basic user information, the dataset includes 12 contextual factors. However, integrating each context variable makes the system more difficult and time-consuming; so, only six contextual variables are chosen. The dataset is divided into 75% for training and 25% for testing. For performance evaluation, this model used three metrics to compute errors in statistics like MAE, MSE, and RMSE and recall, precision, and f-measure for performance analysis. The proposed model has achieved 93.84% precision, 78.54% recall, and 86.90% F-measure.