• clustering algorithms
• data mining techniques
• internet of things (IoT)
• LSTM
• smart devices
• unsupervised learning

This thesis focuses on improving energy management in Internet of Things (IoT) devices by accurately identifying device states to enhance energy conservation and extend battery life. Three frameworks are explored: an unsupervised learning approach using clustering algorithms, an advanced feature extraction method combining Convolutional 1D (Conv1D) and Long Short-Term Memory (LSTM) networks followed by DBSCAN clustering, and a proposed supervised learning approach for future studies. The advanced methods demonstrate significant improvements in clustering quality and state prediction accuracy, contributing to more efficient and sustainable IoT device operations.