AIoT-Based Indoor Air Quality Monitoring and Machine Learning Prediction Framework for Assessing Ventilation Risk in TVET Institutionsutions

Abstract

Inadequate ventilation conditions are commonly observed in Technical and Vocational Education and Training (TVET) institutions, particularly in confined laboratories and high-occupancy indoor spaces where environmental conditions vary with occupancy and operational activities. This study developed and evaluated an Artificial Intelligence of Things (AIoT)-based indoor air quality (IAQ) monitoring and machine learning prediction framework for assessing ventilation risk in TVET institutions. Carbon dioxide concentration, dust levels, temperature, and relative humidity were continuously monitored in selected indoor spaces using an AIoT sensing system. The collected time-series data were processed and used to train a machine learning prediction model implemented using TensorFlow to estimate short-term IAQ trends. Ventilation risk was evaluated by interpreting both observed and predicted values with reference to recognised indoor air quality guidelines. The results showed that short-term prediction enabled earlier identification of potential ventilation risk prior to threshold exceedance, particularly during periods of high occupancy and workshop operation. Compared with real-time monitoring alone, the predictive approach provided earlier warning and improved prioritisation of spaces requiring attention. The findings demonstrate that integrating AIoT monitoring with machine learning-based prediction can support preventive maintenance planning and enhance ventilation management in TVET institutions.