Akademik Veri Yönetim Sistemi
ICHEAS 10TH INTERNATIONAL CONFERENCE ON HEALTH, ENGINEERING AND APPLIED SCIENCES, Dubai, Birleşik Arap Emirlikleri, 1 - 03 Kasım 2025, ss.126-135, (Tam Metin Bildiri)
In recent years, there has been a noticeable increase in the number of small-scale or larger-scale fires worldwide, due to both natural causes and human factors. Fire poses a serious threat that can lead to loss of life and property if not detected early. It also has long-term negative effects on the environment, economy, and society. Traditional fire detection methods are generally based on smoke density and heat measurements. However, these methods can be significantly affected by environmental factors such as wind, humidity, temperature changes, or industrial activities, and therefore do not always provide reliable results. In this study, a smoke detection dataset is used to evaluate the performance of discriminant analysis algorithms in fire detection. The examples in the dataset are divided into train and test sets. Discriminant analysis enables the separation of data into different classes in classification problems and provides a statistical-based approach to better distinguish between fire and normal conditions. Four different models are used: linear discriminant, quadratic discriminant, diagonal linear discriminant, and diagonal quadratic discriminant. The study aims to contribute to the development of fire detection systems by discussing the accuracy, reliability, and practical applicability of the models used. In conclusion, it is observed that discriminant analysis methods have the potential to increase early warning capacity in both indoor and outdoor areas when integrated into existing fire detection systems, thanks to their low-cost and easy-to-implement structure. In this regard, the study makes an important contribution to research and practical applications in the field of fire safety.