OrdinalEncoder and PCA based NB Classification for Leaked Natural Gas Prediction Using IoT based Remote Monitoring System

Abstract

The natural gas (NG), usually methane gas, leaks into the air; it is a big problem for air pollution and the environment. In this paper, we propose to predict gas leakage using ML methods based on the open data provided by the server using IoT-based remote monitoring Picarro gas sensor specification. The performance of the OrdinalEncoder (OE) and MaxAbs normalization-based Naive Bayes techniques was compared with and without the dimensional reduction principal component analysis (PCA) for NG leak prediction. The first step is a preprocessing stage to convert the data based on OE, which results in selecting feature data. The second step is classified into gas CH₄ data by the k-means algorithm. After k-means clustering, the experimental dataset has done an imbalanced data. Therefore, we focusing our proposed models can predict medium and high risk so best. In this case, we compared the receiver operating characteristic (ROC) curve for each classification model. As a result of our experiments, the evaluation measurements include ROC reached 85.3% with the OrdinalEncoder (OE)-NB without PCA for the high-level class; ROC values are 72.2, 73.3, and 75.3% for all classes on the OE_PCA_NB with PCA, respectively. These results showed that the proposed OE-NB and OE-PCA-NB outperformed other models for NG leaks prediction.