A STOCHASTIC MODEL FOR PREDICTING NUMBER OF FIRE ACCIDENT OCCURRENCE IN NIGER STATE

Abstract

Fires are one of the most complex issues that many communities face, as they can cause serious environmental hazards and havoc.Fire outbreaks could be very complicated to quench, yet we cannot totally avoid fire accidents as they (fire) can be ignited from different sources, thereby exposing lives and properties to destruction. The thrust of this research is to provide the government with reliable models to curb the number of fire accidents that occur in order to reduce the loss of lives and property. A stochastic model that predicts the number of fire accident occurrences in Niger State is presented in this thesis. A three-state stochastic model was formulated using the principle of Markov. Each state of the model has four possible observations. The parameters of the model were estimated using the fire accident data collected from the archive of the Niger State Fire Service, after which the model was trained using the Baum-Welch Algorithm to achieve maximum likelihood. The validity test for the model showed 75% accuracy for short-time prediction and 50% accuracy for long-time prediction. This result indicates that the model is more reliable and dependable for short-time prediction. Information for this model could serve as a guide to the government in policy formulation that might assist in curbing the number of fire accident occurrences in the State.