A Hybrid Fuzzy Rough Set, Hierarchical CFA, and Random Forest Approach for Modeling and Validating Voting Intentions: Evidence from the 2023 Thai General Election

Against the backdrop of high digital uncertainty in the 2023 Thai General Election, this study examines how social media reshapes voting intentions through a novel hybrid framework integrating Fuzzy Rough Set Theory (FRST), Hierarchical Confirmatory Factor Analysis (CFA), and Random Forest Regression (RFR). A three-stage design—combining 23 expert opinions with survey data from 812 voters—overcomes expert ambiguity and non-linear dynamics. The findings reveal a hierarchy in digital campaigning: while Party Image (Importance = 0.3056) is the primary predictor for initial voter attention, substantive Campaign Policy (β = 0.98) remains the definitive driver of final commitment. Other perceptual constructs, including Trust, Loyalty, and Perceived Quality, function as reinforcing dimensions that validate policy claims within the digital ecosystem. This suggests a shift where traditional broadcasting is superseded by interactive digital streaming, allowing voters to scrutinize policies through replays and public comments. The model’s robustness, validated through 10-fold Random Forest Cross-Validation, demonstrates high predictive stability (Mean CV R2 = 0.840) and minimal error (MAE = 0.064). This study offers a sensitive instrument for emerging democracies and provides actionable insights, showing that substantive policy remains the ultimate driver of voter choice, even when mediated through Party Image in interactive digital environments.