Time series forecasting using multiplicative model: A predictive model for fire risk in the city of Manila

Abstract

This paper utilized the components in time series such as secular trend, irregular fluctuation, cyclical and seasonal movements to extrapolate prediction using multiplicative model. The main objective of this research is to predict fire risk in short-range period based on the past data. Scrutinizing fire risk based on time series forecasting with multiplicative model, provide possible conclusion on the following questions in analyzing fire risk. Such as, who are the districts that are prone to fire? When is the crucial time that fire will most probably arise? And what is the most possible cause of fire? This study analyzed fire data in the city of Manila from 2011–2015 and predicted the possible occurrence of fire and its risk. A tally of 2823 fire incidents were included in this study and as recorded based on the cumulative frequency, Electrical failure is most cause of fire incident with an average of 315 cases per year. Moreover Tondo is the most densely population in Manila and also predicted as the most risk in fire with 153 possible cases a year prediction with 9.96% absolute mean error. The arises of monthly pattern prediction shows that September is the most risk in fire incidents with possible 61 cases a year, meanwhile summer months are also prone in fire occurrence particularly February, March and May with an average of 57 cases a year with 8.09% absolute mean error which is also the warmest month in a year. Daily pattern show much relationship on the fire incident and fire risk prediction, Saturday is the crucial day in terms of fire risk with 109 possible cases will happen on this day and majority of the possible fire accident will arise between 7:00pm to 9:00pm with 120 possible cases a year with 22% mean absolute error. The significance of understanding fire risk using Time series forecasting must be utilize and administer, which can be useful tool to serve as guide and help to mitigate fire incidents. Future researchers recommend to focus on other variables that contribute fire occurrence to predict fire risk with more accuracy rate.