This study investigates short-term forecasting of indoor air quality in a small industrial environment, using a bakery as a case study. An AutoRegressive Integrated Moving Average (ARIMA) (5,1,0) model was applied to predict variations in carbon dioxide (CO₂) concentrations, a key proxy indicator for ventilation efficiency and occupational exposure, which was monitored over 57 days and modelled to capture short-term fluctuations. The model effectively captured temporal dependencies, with significant autoregressive terms (p 0.05) highlighting the influence of past values on short-term forecasts. However, substantial unexplained variability was observed, suggesting that external drivers such as ventilation, operational schedules, and equipment use play a critical role in shaping indoor air quality. Although ARIMA is a well-established forecasting method, this study demonstrates its value as a low-cost, data-efficient tool for small industries in resource-limited settings where complex monitoring systems and large datasets may be unavailable. By forecasting CO₂ dynamics, facility operators can anticipate periods of elevated indoor exposure and adjust ventilation strategies to mitigate risks, thereby supporting worker health, safety, and productivity. The findings provide a proof-of-concept for integrating classical time-series forecasting into occupational environmental management. However, future research should extend this approach to multivariate or hybrid ARIMA–machine learning models, incorporate additional pollutants with direct public health implications e.g., PMs, NO₂, temperature, and relative humidity, then validate across multiple industrial contexts. This would enhance generalizability and strengthen the role of forecasting tools in informing ventilation policy, exposure mitigation, and regulatory compliance in small-scale industrial environments.
Olatona et al. (Thu,) studied this question.