Water Pump Health Prediction in Industrial Wireless Sensor Network Using Supervised Learning

Nowadays, wireless sensor networks (WSNs) are often used in industrial settings to gather data regarding the different machines in use. There are numerous sensors such as pressure gauges, temperature sensors, dust sensors, etc. that are used to gather data in real-time. These sensors pass on the information to the base station from where it is further sent to the server. The data collected are then analyzed to determine the working condition of the different machines used in industry. The data obtained via the sensors are huge and cannot be manually examined by a person. Rather the data is fed to some machine intelligence (MI) systems which analyze the data and derive suggestions based on prior learning experience. In this work, an MI-based approach is used for the prediction of water pump risk in industrial WSN. This work is focused on several supervised machine learning (ML) based models such as Neural Network (NN), Decision Tree (DT), Support Vector Machine (SVM), and Naive Bayes (NB) to perform predictions. This work is carried out using cross validation by considering the number of folds (NFD) as 2, 3, 5, and 10. These models are evaluated using the classification accuracy (CA) performance metric. This work is implemented using Python-based Orange 3.26.0.