KEPUTUSAN MANUSIA VS KEPUTUSAN MESIN: STUDI KOMPARATIF TERHADAP AKURASI DAN KONSISTENSI DALAM PENGAMBILAN KEPUTUSAN

Abstract

This study aims to analyze and compare the accuracy, consistency, and decision-making efficiency between humans and machine learning (ML) algorithms in tabular data classification tasks. The dataset comprises 50 classification cases containing both numerical and categorical features with binary decision labels. Two groups were compared: 10 human participants, and six ML algorithms—Logistic Regression, Support Vector Machine, Random Forest, Decision Tree, k-Nearest Neighbors, and Naive Bayes. ML models were trained on 80% of the data and tested on the remaining 20%, while human participants manually classified all 50 test cases. The results showed that the average human accuracy was 76.2%, while ML algorithms achieved between 78.9% and 91.8%, with Random Forest yielding the highest performance. Human decision-making took an average of 18 seconds per case, significantly slower than the algorithmic predictions completed within milliseconds. Additionally, high variability in human responses indicated lower consistency compared to deterministic outputs from ML models. These findings support the integration of ML algorithms as a decision support or replacement tool in data-driven domains, with the potential to reduce human error in high-stakes environments. Nevertheless, human involvement remains essential in contexts requiring ethical consideration and interpretability.