Integrating Enhanced Learning to Rank into a Hybrid Deep Learning System for Optimized Recommendations
Naveen Kumar Navuri, C. V. P. R. Prasad · 2025
We present a state-of-the-art LTR method that is integrated into our recommender system to boost its performance. We’re starting with a recommendation list generated by a custom-built deep-learning framework. For a better ranking of these items, we use a listwise LTR method specifically designed for this purpose. This LTR model is trained for optimizing the key performance metrics like Normalized Discounted Cumulative Gain (NDCG), and Mean Average Precision (MAP) which are important for top-N-based evaluation. Our experimental results mainly conducted on the Movie Lens dataset show that the proposed system significantly outperforms conventional models in ranking performance.