Nowadays the Psychological stress is becoming a threat to people’s health, as the time is moving stress level is increasing so quickly. With the rapid pace of life, more and more people are feeling stressed. In the proposed method, we will find the stress state of society very closely related to their relatives or friends in social media and will give work to a large-scale dataset from the real-world social platforms to study the correlation of user's stress states and social interactions. As the peoples are sharing their day to day activities on social media platforms or the interaction between the peoples makes social media so popular in these past few years. The social media platform makes it convenient to hold all the online social network data for stress detection, though the stress itself is a non-clinical and common in our day to day lifestyle. Excessive and chronic stress can be rather harmful to people’s physical or mental health. With the development of social networks like twitter, Whatsapp and Facebook, more and more people are willing to share their daily activities and moods with friends through the social media platform. In this, we will first define a set of stress-related textual, visual, and social attributes from various aspects, and after that, we will propose a novel hybrid Model Stress Detection Model (SDM). This proposed hybrid model compression. In this by further analyzing the social interaction in the data, we will also discover the different several fascinating phenomena, i.e. the number of the social structures with no delta connections of the stressed users is around 38% higher than that of the non-stressed users, recommending that the social structure of the stressed users tends to be less connected and less complicated than that of the non-stressed users.