A Collaborative Filtering Recommender System using Apache Mahout, Ontology and Dimensionality Reduction Technique

Making improvements in the efficacy of methodology incorporated in the recommender engine is the biggest challenge. It's also vital to think about achieving a balance between two desirable property accuracy and time to find generate recommendation item set when promoting goods via recommender engine because it needs to generate correct suggestions in real-time. This study provides a novel approach for recommendation engine based on Collaborative Filtering (CF) methodologies in this respect. As a result, we use dimensionality reduction approach of linear algebra and ontology approach for semantic similarity to address two major problems in recommendation engines: 1. sparsity and 2. scalability. We incorporated ontology to make improvements in recommendation efficacy and used Singular- Value factorization approach to handle dataset-sparsity and to better scalability of recommendation engine technique. We demonstrate the method's efficiency by evaluating it on real-world Movie-Lens dataset and comparing the findings to those of other approaches in the literature. The solution proposed by us improved the sparsity and scalability concerns in CF, according to the results.