Development of SMOTET-LSTM Model Based on Hyperparameter Tuning for Fault Classification in Multi-Sensor Nodes

In the context of Internet of Things (IoT) structures, sensor nodes have been observed to generate erroneous data due to their constrained operational capacity and position. The presence of faulty nodes can lead to significant challenges in communication, data traffic, and data evaluation. Consequently, it is imperative to segregate data obtained from faulty nodes from standard data. Concurrently, the identification of the specific fault type is paramount. The present study utilised machine learning and deep learning techniques to classify fault types, with the data collected from 54 sensors in a closed building over a period of 3 months. Initially, the performance analysis of the LSTM model was compared with that of Support Vector Machine (SVM), K-Nearest Neighbor (KNN) and Random Forest (RF) machine learning algorithms, given the utilisation of large amounts of data. Subsequently, as certain classes were characterised by limited data, data augmentation was implemented using synthetic data, and the SMOTET-LSTM model was developed through HPO (Hyper Parameter Optimization). This model demonstrated superior performance in comparison to the other algorithms.