Recently, social media platforms have been widely used as a communication tool on social networks. Many users have utilized these platforms to reflect their personal lives. These users differ in terms of background, language, age, and educational level. The close relationship between these platforms and their users has created rich information that is related to these users and can be exploited by researchers. Their posts can be analysed using natural language processing (NLP) to predict psychological traits such as depression. However, to the best of our knowledge, no study has utilized social media to predict mental health disorders in Arabic posts, especially depression. Therefore, in this study, we investigate the application of natural language processing and machine learning on Arabic text for the prediction of depression, and we evaluate and compare the performance. Our research method is based on the collection of Arabic text from online forums and the application of either a lexicon-based approach or a machine-learning-based approach. In the former approach, the ArabDep lexicon is created, and a rule-based algorithm is used to predict depression symptoms using the created lexicon; however, in the latter approach, the data are annotated with the help of a psychologist, text features are extracted from Arabic posts, and machine learning algorithms are ultimately applied to predict depression symptoms. We demonstrate that our applied approaches exhibit promising performance in predicting whether a post corresponds to depression symptoms, with an accuracy of more than 80%, a recall of 82% and a precision of 79%.