Abstract
This study examines the comparative performance of different data types in machine learning and artificial intelligence applications using gray relational analysis (GRA). The research evaluates five distinct data types across four key machine learning tasks—image, spectrogram, photometry, light curve, and time series—for classification, regression, clustering, and prediction. By systematically applying the GRA method, the study transforms multi-objective optimization problems into single-objective evaluations to identify the most versatile data types for AI applications. Photometry performs best overall, achieving the highest gray relational grade (GRG) of 0.7876, with particularly strong results in classification (98.60) and regression (80.00). Image data comes in second with a GRG of 0.7684, excels in clustering (23.00), and provides solid prediction results (40.00). Spectroscopy data ranks third (GRG: 0.6602), while light bending and time series data rank fourth and fifth, respectively. While the overall GRG (0.3864) is very low, time series data shows significant strengths in forecasting (70.67) and clustering (65.00) tasks. This research advances the knowledge of data-driven decision-making in AI applications across a variety of domains that require optimization across multiple performance metrics.