<![CDATA[Sex identification from human skulls is a crucial aspect of forensic anthropology; however, traditional methods still face limitations such as subjective assessment and inter-population variation. This study proposes the application of Information Gain as a feature selection technique and Random Forest as a classification algorithm for sex determination based on craniometric data. The dataset used is the Howells dataset consisting of 2,524 samples with 83 skull measurement features. Feature selection using Information Gain was performed with threshold values of 0.01, 0.05, and 0.09, followed by additional testing across a threshold range of 0.01 to 0.09. Model evaluation was conducted using 10-Fold Cross Validation with default Random Forest parameters. The results show that a threshold of 0.02 produced 57 selected features from the original 83, achieving the best performance with an accuracy of 87.40%, precision of 87.53%, recall of 87.40%, and F1-score of 87.41%. These results outperform the baseline model without feature selection, which achieved an accuracy of 86.57%. This study demonstrates that Information Gain feature selection can reduce data dimensionality by 31.3% while simultaneously improving sex classification performance based on craniometric data.]]>
Copyrights © 2026
