Accurate long-term time series forecasting is essential across various domains, but it requires understanding interdependencies among variables. Data-driven deep learning methods struggle with capturing long-term dependencies effectively. This paper proposes a hybrid model that combines Sparse Identification (SI) with Convolutional Neural Networks (CNN) to enhance interpretability and generalization in forecasting. The SI method extracts trends, seasonality, and periodicity, followed by deep neural networks capturing intricate relationships. Experimental results demonstrate the model’s high accuracy and practicality, contributing to advancements in time series prediction methodologies.