Super typhoons can lead to post-traumatic stress disorder (PTSD), which can adversely affect a person's mental health after a disaster. Neuroimaging studies suggest that patients with PTSD may have post-exposure abnormalities of the white matter. However, little is known about these defects, if they are localized to specific regions of the white matter fibers, or whether they may be potential biomarkers for PTSD. Typhoon survivors with PTSD (n = 27), trauma-exposed controls (TEC) (n = 33), and healthy controls (HCs) (n = 30) were enrolled. We used automated fiber quantification (AFQ) to process the participants' DTI and compared diffusion metrics among the three groups. To evaluate diagnostic value, we used support vector machine (SVM) and a random forest (RF) classifier to build a machine learning model. White matter fiber segmentation between the three groups was found to be statistically significant for the fractional anisotropy (FA) value of the right anterior thalamic radiation (ATR) (26-50 nodes) and right uncinate fasciculus (UF) (60-72 nodes) (FDR correction, p < 0.05). By analyzing the characteristics of the machine learning model, the two most important variables were the right ATR and right UF for differentiating PTSD and trauma-exposed controls (TEC) from the healthy controls (HC). In addition, the left cingulum cingulate and left UF were the most critical variables in the differentiation of PTSD and TEC. AFQ with machine learning can localize abnormalities in specific regions of white matter fibers. These regions may be used as a diagnostic biomarker for PTSD.