MULTI-OBJECTIVE MODEL SELECTION IN TIME SERIES FORECASTING: A GENETIC ALGORITHM-BASED MATHEMATICAL APPROACH

Abstract

Time series forecasting plays a critical role in various domains such as finance, weather prediction, and supply chain management. Selecting an optimal forecasting model involves balancing multiple conflicting objectives, including prediction accuracy, model complexity, and computational efficiency. This paper proposes a novel multi-objective model selection framework that leverages a genetic algorithm (GA) to simultaneously optimize these criteria. By encoding candidate forecasting models as chromosomes, the GA explores the solution space to identify Pareto-optimal models that offer the best trade-offs among accuracy, complexity, and run-time. Experimental results on benchmark time series datasets demonstrate the effectiveness of the proposed approach in improving forecast quality while maintaining model interpretability and computational feasibility. This method provides a robust tool for decision-makers to select forecasting models that align with diverse operational requirements.