<p>Burnout is a common phenomenon and appears to be comprehensive in all walks of life. In recent years, the World Health Organization (WHO) has included it in the International Classification of Diseases. The growing discussion by international organizations shows increasing public concern about the impact. In addition, more related studies have pointed out that it has caused a significant economic burden on the world. At this stage, the judgment of burnout is still vague and can only rely on the subjective judgment of the relevant scales, but it is still difficult to be a reliable indicator. An objective and effective burnout assessment system is crucial. This study chose the police as the starting point. Using functional near-infrared spectroscopy (fNIRS) to measure changes in hemoglobin concentration in the prefrontal lobe during 33 active police officers performing the mental arithmetic task and the verbal fluency task. We extracted the features after processing the signals. After that, the group was marked with the relevant scale score and then imported into the support vector machine algorithm. Finally, a model with a training accuracy rate of 91.3% and a test accuracy rate of 90.0% was obtained, demonstrating that the burnout assessment system using fNIRS combined with machine learning is promising.</p>