<![CDATA[This research implements a Long Short-Term Memory (LSTM) algorithm to predict temperature patterns as part of an IoT-based monitoring system designed to mitigate fire risks. The dataset is collected from ESP32-based sensors that record temperature and timestamp data. The LSTM model was trained using normalized temperature data, with five-step ahead predictions. Preprocessing included combining date and time into a datetime index, followed by scaling and reshaping data for supervised learning. The model architecture consists of a single LSTM layer and a dense output layer. The prediction results show a low Mean Squared Error (MSE), indicating the LSTM model is effective for early detection of potential fire hazards. This work contributes to real-time risk mitigation by improving the predictive accuracy in IoT environments.]]>
Copyrights © 2025
