Hair and fingernails, emerging as crucial biological physical evidence, play a prominent role in geographical traceability studies within the forensic anthropology. This study explored the potential of terahertz spectroscopy and chemometrics techniques in identifying hair and fingernail samples from various geographical regions. A total of 196 hair and fingernail samples underwent scanning with terahertz time-domain spectroscopy. Different chemometric tools, including multilayer perceptron, support vector machine, and Fisher discriminant analysis, were employed for model building on fingernail dataset, hair dataset, and fusion data, respectively. The performance of models constructed based on different datasets was compared and evaluated using metrics such as accuracy, precision, recall, and F-score. Results revealed that the fusion data, employing the Fisher discriminant analysis (FDA) algorithm, exhibited the best recognition ability, achieving a 100 % accuracy rate and F-value. Terahertz spectroscopy combined with FDA algorithm emerged as an effective and convenient tool for differentiating hair and fingernail samples from diverse geographical regions. This innovative technique boasts environmental friendliness, rapid analyses, and minimal sample preparation requirements compared to traditional methods.