Recently, research has been conducted on the use of social networks as a new platform for identifying people with mental disorders. In addition, because of the complexity of diagnosing psychological diseases using conventional methods, the use of machine learning for identifying theses psychological diseases is increasing. The goal of this article was to systematically review the research conducted using social media data for predicting and diagnosing psychological disorders with the help of machine learning. Based on systematic review on the Prisma method, the aim of this article was achieved through searching the main keywords of diagnosis and the prediction of mental disorders combined with machine learning and social media data without considering the dates of their publications. Depression had the highest frequency among the final 20 selected articles with a predictive power of 42% and 87%, the lowest and the highest respectively. On the other hand, only 30% of studies used questionnaires for gathering data on social media and the most common approach for data collection was public posts on social media by the use of regular expressions. Twitter has also been used as the largest source of data collection in these sorts of studies. It seems that computational psychology based on machine learning methods could help to identify disorders at an appropriate time and select more effective treatments for mental disorders among the users of social media.