ABSTRACT
Disasters often lead to economic loss and loss of human life. Early predictions corresponding to disasters can allow administration to take preventive and precautionary measures. Forest fires are common and uncertain disasters that can occur due to disturbances in normal temperature stability.
Forest fires are often accompanied by earthquakes, rain, or eruptions. The entire framework associated with the proposed system consists of a sensor, fog, and a cloud layer. Data acquisition within the sensor layer collects the data on the soil and land through the sensors.
Furthermore, pre-processing is performed at the sensor layer. The pre-processing mechanism removes noise from the dataset. The fog layer contains a feature reduction mechanism that is used to reduce the size of the data to conserve the energy of sensors during the transmission of data.
Moreover, predictor variables selected within the energy conservation mechanism are used for exploratory data analysis (EDA). The main characteristics of the data are extracted using EDA.
Furthermore, principal component analysis applied to the fog layer analyses the dependencies between the attributes. Dependencies are calculated using correlation. Negatively skewed attributes are rejected; thus, the dimensionality of the dataset is reduced further.
All gathered prime attributes are stored within the cloud layer. K-means clustering is applied to group similar entities within the same cluster. This study proposes an early warning system for forest fires.
