Prediction of the toilet’s health and manpower deployment orchestration

Abstract

In the era of technology-driven innovation, the Internet of Things (IoT) has greatly influenced many areas of our lives. Utilisation of the IoT provides an opportunity to develop Smart Toilets, where washrooms are embedded with sensors to detect different aspects of the facilities, providing real-time statistics about the environment. In this thesis, we investigate the possibilities of using predictive analytics to uncover real-time insights about the washrooms’ hygiene and predict future usage and conditions of the facilities. The purpose is to improve the reliability, availability, and maintainability of the washrooms through the optimisation of cleaning schedule based on predictive analysis. An efficient deployment of manpower for cleaning and maintenance allows the upkeep of washroom hygiene standards and quality of service while reducing manpower wastage and conserving resources. The proposed approach includes using descriptive analysis and data exploration to determine trends, patterns to better generate hypotheses regarding the raw sample data, followed by data treatment to handle errors and impute missing values using Python. To model the relationships and dependencies between prediction output and input, different supervised machine learning algorithms including Extreme Gradient Boosting (XGBoost) and Random Tree. From the findings, conclusive relationships can be determined between the different inputs and outputs and the results are crucial for future studies where more rigorous machine learning models can be developed for better performance.