Depression is a major mental health issue which adversely impacts an individual’s health and wellbeing. Many people experience depression but often hesitate and do not want to talk about it, which can lead to bigger problems like feeling isolated, facing discrimination, or even self harm. With the rise of social media, people are now sharing their thoughts and feelings more openly. This study presents a novel approach to identifying depression by analyzing code-mixed language on social media platforms, leveraging advanced natural language processing (NLP) techniques. TF-IDF and Word2Vec word embeddings are used for extracting important features. Machine learning models, such as CatBoost, XGBoost, GaussianNB and AdaBoost are then utilized to improve accuracy. The CatBoost model achieves the highest accuracy of 0.95, outperforming traditional methods. The results demonstrate the potential for using social media data to detect depression. This research highlights the need for effective methods that work well with multilingual and mixed-language content on online platforms.