Application of random forest method for analysis of gas sensor readouts from mold-threatened buildings

• Main Authors: Łagód, Grzegorz, Majerek, Dariusz, Kočí, Vaclav, Černý, Robert
• Format: Conference Proceeding
• Language: English
• Subjects: Decision trees; Gas sensors; Mold growths; Sensor arrays; Sensors; Sick building syndrome; Supervised learning; System effectiveness
• ISSN: 0094-243X; 1551-7616
• DOI: 10.1063/5.0147200

Humidity increase in buildings frequently leads to the growth of mold, which is one of significant factors for evaluation of Sick Building Syndrome. The assessment of mold contamination level in buildings based on gas sensors array readouts is considered as a cheap and fast detection technique; nonetheless, interpretation of signals is quite complex, mostly because the signals obtained from sensors are multidimensional. Furthermore, there is no sole reference method used in practice. The signals analyzed in the original multi-dimensional space are characterized by high variability, depending on the conditions in the tested buildings. In such a situation, the random forest methodology can be applied, which till now has been successfully involved to a wide range of prediction problems and has few parameters to tune. Aside from being simple to use, the method is generally recognized for its accuracy and ability to deal with small sample sizes and high-dimensional feature spaces. At the same time, it is easily parallelizable and therefore has the potential to deal with large real-life systems. This supervised learning procedure operates according to the simple but effective “divide and conquer” principle: sample fractions of the data, grow a randomized tree predictor on each small piece, then paste (aggregate) these predictors together.