Depression is one of the most common diseases these days due to many economic and financial problems that befell individuals and families, especially in Nigeria. It is the major reason for anxiety and sleeping disorders. It can have a significant impact on a person's daily functioning and their ability to participate in their work, school, meals, sleep and play. Depression, if left untreated or not given early attention, may sometimes lead to self-harm and suicide. The study evaluates the effectiveness of SVM, RF, DT, K-NN, and NB machine learning models. It proposes a hybrid SVM-RF model to detect likely depressed posts on Facebook and X (Twitter) using users' tweets and comments. Datasets were collected via Facebook using the Facebook_scraper library and the Sentiment140 dataset on the Kaggle website. Datasets were split into 80:20 for model training and testing. Python 3.7 and Google Colab were used to analyse the machine learning algorithms. This study showed that SVM outperformed other techniques, such as RF, DT, K-NN, and NB, achieving 91% accuracy in identifying depressive content. The outcome also shows that the hybrid SVM-RF model achieves 93% better performance in capturing and understanding more complex patterns of depression-related language, as measured across several metrics. The outcome of the study will help psychologists, policymakers, and other concerned stakeholders in society, as well as the Nigerian government, to identify vulnerable individuals who are at risk of experiencing depression among social media users.