A Machine Learning-Based Disinfectant Type, Concentration, and Usage Monitoring System for Real-World Scenarios

Different types of disinfectants are used to maintain the hygiene of workplaces and houses. Due to negligence of the office/house workers, improper and irregular application of such disinfectants can cause severe health hazards. Therefore, it is important to ensure that recommended disinfectants are applied timely and per the procedure. The world has witnessed the importance of sanitation during the recent pandemic (COVID-19) as well. In this work, we have proposed a non-selective six MQ series elements-based gas sensor array followed by analytics using machine learning to monitor the type, concentration, and application of six different types of disinfectants, commonly used in real-world scenarios. The experimental dataset was obtained by exposing the sensor node prototype with the considered disinfectants (15 ml of each) for 15 minutes, taken one by one. Multiple machine learning models were then trained on this dataset to classify unseen data samples with high accuracy. Decision Tree (DT), Random Forest (RF), and Gradient Boosting (GB) models all reached 100% classification accuracy, but models based on Support Vector Machine (SVM) and Logistic Regression (LR) could only manage 99.29% and 98.70% accuracy, respectively. Using Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and R2 Score, the classification performance error was also calculated. The DT, GB, and RF have the minimum possible error with the highest R2 Score with a value of 1.00, while LR and SVM have nominal errors with R2 Scores of 0.984 and 0.975, respectively. It is demonstrated that our proposed system can detect the application of various disinfectants in real-world situations with very high accuracy.