ATMOSPHERIC SMOKE DETECTION SYSTEM
ARUN PRASATH I
M.Sc. (Decision and computing Sciences) – IVth year
Coimbatore Institute of Technology
Coimbatore, India
1933004mdcs@cit.edu.in
CHANDIA S
Assistant Professor
Dept. of Computing (DCS)
Coimbatore institute of technology
Coimbatore, India
schandia@cit.edu.in
GOWSIKKAN S
M.Sc. (Decision and computing Sciences) – IVth year
Coimbatore Institute of Technology
Coimbatore, India
1933014mdcs@cit.edu.in
HARISH S
M.Sc. (Decision and computing Sciences) – IVth year
Coimbatore Institute of Technology
Coimbatore, India
1933017mdcs@cit.edu.in
Abstract—The detection of smoke is a critical task in various applications, including fire detection and air pollution monitoring. Traditional smoke detection systems often rely on sensors that detect changes in temperature, humidity, and light, but these methods can be slow and prone to false alarms. In recent years, machine learning algorithms have shown promise in improving the accuracy and speed of smoke detection systems. Decision tree algorithms, in particular, are known for their ability to handle complex datasets and produce interpretable models, making them a promising approach for smoke detection. In this research, we investigate the use of decision tree algorithms for smoke detection and evaluate their performance in comparison to other machine learning algorithms. We use a dataset of smoke images and preprocess the data using feature extraction, normalization, and dimensionality reduction techniques. We train and evaluate decision tree models using different hyperparameters and feature selection techniques and compare their performance with other machine learning algorithms. Our experimental results show that the decision tree algorithm outperforms other machine learning algorithms in terms of accuracy and interpretability. We also show that feature selection and pruning techniques can further improve the performance of the decision tree model. Our research demonstrates the potential of decision tree algorithms for smoke detection and provides insights into their strengths and limitations.
Keywords—machine learning, random forest, prediction, decision tree.