An external stimulus, event, or environment that stresses an individual is called a stressor. Many mental stress detection studies have been focused on the discrimination of the mental state with and without the experimental stressor. However, the mental state in the absence of experimental stressors may not represent accurately the nonstress (baseline) state because people inherently experience considerable stress in their daily lives. Therefore, we assumed that stress detection could be improved more accurately by considering the daily stress. In this study, functional near-infrared spectroscopy (fNIRS) was measured in 41 healthy participants to quantify their prefrontal cortical oxygenation during performing a cognitive task as an experimental stressor, considering individual daily stress level based on the self-report. We then extracted six signal features, including the slope, mean, standard deviation, peak, skewness, and kurtosis of the oxygenated hemoglobin concentration. Using various feature combinations and time windows, we successfully managed to classify daily stress (high/low) and the mental state (task/rest) with support vector machine classifiers. Specifically, individual daily stress level can be easily discriminated with signal features from fNIRS. Moreover, mental state classification performance improved significantly when the daily stress level was handled separately. The findings of this study show the feasibility of the fNIRS-based daily stress classification and can be used in the future to design a robust mental stress management system for the assessment of daily stress in individuals.INDEX TERMS daily stress, functional near-infrared spectroscopy, stroop word color task, mental stress classification.