Stress is a major factor for human health in todays fast paced world. It occurs due to work burden, life problems, responsibilities. It is a human response to ventures. This can result in various reactions in the life. The detection of stress at an early stage can be beneficial for prevention of various body problems like nervousness, unhappiness, cardiovascular diseases, and other stress related complications. By analyzing physiological signals which are an indication of response of stress in the body, we can save a human being. By observing heart rate variability, heart activity and various signals like Electrocardiogram (ECG) and other conventional methods detect stress by the electrical activity of the heart. This review describes the favorable and unfavorable methods of Machine Learning (ML) based models. These models provide precise and immediate evaluations of stress levels by using physiological data. These studies report the stress levels from low to high levels, which show the present growth and advancement of these models.