In the context of smart classroom environments, this study proposed an intelligent control framework for automating the operation of IoT devices based on real-time environmental data collected from sensors via an ESP32 microcontroller. The framework was designed to enhance classroom comfort by dynamically adjusting devices such as lights, fans, and air conditioners in response to changing conditions, thereby minimizing manual intervention and improving the overall teaching and learning experience. To evaluate the effectiveness of this approach, several machine learning and deep learning models-namely Random Forest, XGBoost, Long Short-Term Memory (LSTM), and Gated Recurrent Unit (GRU)-were implemented and compared. Experimental results showed that LSTM and GRU achieved significantly higher prediction accuracy, up to 99%, compared to traditional approaches. However, this improvement came at the cost of increased computational time during training and inference. Despite this trade-off, the deep learning models demonstrated strong potential for intelligent, data-driven control in dynamic classroom environments.