Integrating MCDM techniques for optimized recommender systems in the age of generative AI

With the increasing adoption of artificial intelligence in various domains, this research proposes an innovative approach to enhance the performance of recommendation systems by integrating multi-criteria decision-making (MCDM) techniques with machine learning algorithms. The study addresses key challenges in recommendation systems, including multi-criteria preferences, cold-start problems, sparse data, and recommendation diversity, using the Amazon dataset. The methodology implemented in this work combines MCDM approaches with collaborative filtering (CF) based on singular value decomposition (SVD) and content-based filtering (CBF) models. The primary focus of the study is to evaluate the stability, generalization capability, and influence of MCDM on performance metrics, such as root mean square error (RMSE) and mean absolute error (MAE). The analysis results indicate that the SVD-CF model demonstrated stability in the validation phase but experienced overfitting when applied to new data, revealing limitations in pattern generalization. In contrast, the CBF model demonstrated superior generalization capabilities and produced more accurate predictions for unseen data. Although MCDM had an effect on RMSE and MAE in both models, this method showed promise in improving performance, particularly in models that integrate current data or are combined with other techniques. These findings suggest that MCDM can contribute to reducing overfitting and improving the accuracy of new data. The consistent performance of the CBF model affirms its reliability in recommendation systems. Further research is recommended to focus on the deeper integration between MCDM and advanced machine learning techniques to enhance recommendation accuracy, adaptability, and user satisfaction.