What question did this study set out to answer?

July 11, 2026PeerJ1 citationsOpen Access

Application of machine learning algorithms in building health diagnostics: predictive analytics evaluating indoor air quality and sick building syndrome in educational settings

SISyazwan Aizat IsmailUniversiti Sains Malaysia HKHaslinda Mohamed KamarUniversity of Technology Malaysia NKNazri KamsahUniversity of Technology Malaysia

Key Points

To develop a predictive analytics model for diagnosing sick building syndrome using indoor air quality data.
Collected health complaint data via a modified questionnaire and checklist.
Conducted indoor air quality assessments using occupational hygiene techniques.
Utilized machine learning algorithms including Neural Network and Random Forest to analyze the data.
Neural Network achieved a classification accuracy of 73.8%.
Formaldehyde and Total Volatile Organic Compounds were identified as top predictors for health complaints.
Strong associations were found between multiple indoor air quality parameters and health complaints, especially in mechanically ventilated environments.

Abstract

Our research aimed to develop and validate a predictive analytics model for diagnosing sick building syndrome (SBS) in learners. We achieved this by gathering and analyzing epidemiological and exposure assessment data. The current assessment involved the use of the modified MM040NA SBS questionnaire and checklist from Indoor Air Quality (IAQ) Industry Code of Practice (IAQ-ICOP) from Department Occupational Safety and Health (DOSH), Malaysia, with participants scoring of their answers, recording and scoring of their simultaneous self-reported and physician-ascertained health complaints. At the same time, IAQ assessments were collected at the location of each participant with the use of occupational hygiene techniques. Several predictive analytics algorithms, namely Neural Network, Logistic Regression, Classification Tree, Random Forest, and Support Vector Machine, were used to train and test the collected data set. The Neural Network model rendered the most effective classification accuracy, reaching 73.8%. Validation also showed that multiple IAQ parameters were strongly associated with health complaints, especially in mechanically ventilated environments. Variable importance analysis identified Formaldehyde and Total Volatile Organic Compounds (TVOCs) as the top predictors for health complaints, highlighting their critical role in indoor environmental quality. Taken together, the results confirm the effectiveness of neural network-based predictive analytics in correctly diagnosing sick building syndrome (SBS) and related health complaints on limited IAQ data and thereby improving the ability to assess during the early stages.

问 AI

Bookmark

View Full Paper