Expert System for Early Childhood Talent Detection Using Certainty Factor and Dempster Shafer Algorithms

Abstract

Early life is a crucial window for recognizing children’s interests and talents that shape later development. This study implements and compares two reasoning algorithms—Certainty Factor (CF) and Dempster–Shafer Theory (DST)—within a rule-based expert system designed to determine early-childhood interests and talents. Observable “symptoms” (behavior, preferences, and responses to stimuli) are mapped to potential talents, including linguistic, musical, logical-mathematical, and kinesthetic intelligences. The CF module computes confidence values from expert-assigned belief weights, yielding a single interpretable score per talent; the DST module aggregates evidence while explicitly representing uncertainty through basic probability assignments over the frame of discernment. We evaluate both methods in the deployed application with respect to accuracy, decision consistency, and response speed. Results show that, for the representative trait set aligning with linguistic indicators, CF produced the highest agreement with expert judgment 84% confidence while DST assigned 65% mass to the same singleton hypothesis, reserving the remainder for competing hypotheses and ignorance. These findings indicate that CF offers a more decisive signal under congruent evidence, whereas DST contributes caution by quantifying residual uncertainty. Together, the dual approach supports transparent and scalable screening of early talents, enabling caregivers and educators to act when support is strong and seek additional observations when uncertainty persists.