Accurate prediction of foreign exchange (forex) rates is a critical task for financial markets, influencing decision-making for traders, investors, and policymakers. Traditional time series forecasting models often have difficulty obtaining the complex trends and non-linear dependencies inherent in forex data. To tackle these challenges this study comes up with a combined model that incorporates the powerful qualities of Prophet and LSTM networks. Prophet is leveraged for its ability to model trend and seasonality in time series data while handling irregularities and missing values effectively. In order to capture the residual non-linear dependencies left by Prophet, we integrate an LSTM network, recognized for its capability for modeling long-term dependencies in data that are sequential. The hybrid model obtained an MAE of 0.00051, an RMSE of 0.00064, and a MSE of 0.00375, outperforming traditional models and highlighting the effectiveness of this approach in improving forex rate predictions. This study not only advances time series forecasting techniques but also offers valuable insights for stakeholders in the forex market, enhancing their ability to make informed decisions.