INTEGRATED METHOD FOR SYNTHESIS OF RECOMMENDATIONS BASED ON MULTI-CRITERION ANALYSIS AND COLLABORATIVE FILTRATION

Abstract

 The article examines a novel approach to personalized recommendations based on a combination of multi-criteria analysis, dynamic weighting, loss-gain estimation, and collaborative filtering. This method addresses the problem of optimizing decisions in scenarios with conflicting criteria by balancing objective factors such as price or quality with the subjective preferences of users. By dynamically adjusting the weights of criteria using behavioral data, this approach adapts to changing user priorities and contexts, providing relevant and accurate recommendations.

The main innovation is the integration of Pareto analysis to narrow down the optimal options, quantify trade-offs using gain-loss estimation, and refine the results using collaborative filtering that uses feedback from users with similar preferences. These components are combined into a single framework to provide a comprehensive and adaptive decision-making system. The practical application of the method is demonstrated on the example of residential real estate selection, where it effectively combines criteria such as price per square meter, developer rating, and location preference.

The proposed system not only improves recommendation personalization, but also offers scalability for various domains, including e-commerce, real estate, and asset management. Its adaptability is achieved by incorporating both quantitative assessments and qualitative insights derived from user behavior.

Future research directions include extending the method to handle larger datasets with higher dimensional criteria, integrating advanced machine learning models for dynamic parameter optimization, and improving computational efficiency for real-time implementation. This framework has significant potential for solving complex decision-making tasks in both individual and organizational contexts, facilitating the development of more sophisticated and user-centric recommendation systems.