Optimize cleaning school’s restroom by WSN and LSTM approach

Abstract

The detection and prediction of cleaning conditions in school restrooms are crucial for reducing health risks and improving service quality. Traditional methods like manual hygienic inspection, fixed cleaning schedules, and automatic flushing devices have required large investments of money and effort from cleaning businesses to maintain cleanliness in school restrooms. To address this issue, we propose a prediction model based on Recurrent Neural Network (RNN) and Long Short-Term Memory (LSTM) architecture. The model uses a dataset obtained from real-time conditions of the toilet via a wireless sensor network, enabling more efficient scheduling of toilet cleaning tasks. By predicting patterns of Ammoniac (NH3) concentrations and Relative Humidity (RH) levels over time, our LSTM model is superior to the RNN model in performance, significantly reducing deviations in the NH3 and RH values with RMSE values of 3.32 and 2.85, respectively. Furthermore, the model’s flexibility allows a variety of inputs to evaluate the need for cleaning at specific times, achieving maximum efficiency without requiring excessive neurons.