Social media trends can be predicted using a variety of advanced machine learning and deep learning models, including LSTMs (Long Short-Term Memory Networks), Temporal Convolutional Networks (TCNs), GRUs (Gated Recurrent Units), Transformers, and XGBoost, an optimized gradient boosting algorithm. To identify relevant keywords, we used Natural Language Processing techniques and then collected time series data from Google Trends using web scraping techniques, tracking the popularity of specific keywords over a period of time. These models are then used to forecast the trajectory of trending topics, how quickly they grow and when they might decline. The goal of this paper is to achieve successful and reliable predictions of which topics will gain popularity, how quickly they will grow, and when they might begin to decline, offering valuable insights for social media analysts and marketers. GRU combined with XGBoost achieved the lowest mean squared error loss of 0.0106, outperforming all other models.