A Forecasting-Based ARIMA Modeling of Cancer Deaths in Bangladesh and World

Faruq Abdulla

Department of Statistics, Islamic University, Kushtia-7003, Bangladesh.

E-mail: faruqiustat09mnil@gmail.com

Abstract

Till now, cancer is concerned as the pandemic second leading cause of deaths and day-by-day it increasingly threatens human life. In 2017, a total of 9.56 million and 99,302 peoples have been passed away from this world and Bangladesh, respectively by cancer. In any research field, the accuracy of the future prediction based on previous information is highly competitive as well as challenging task and the future direction will be appreciated as well as important to the policymakers. Addressing this issue, the author of this research attempts to identify the behavior of the cancer deaths by the Auto-Regressive Integrated Moving Average (ARIMA) model that could be used to forecast the number of deaths due to cancer. This study considered secondary yearly data of overall, age-specific, and cancer-type-specific deaths over the period 1990 to 2017 in Bangladesh and globally. The best ARIMA model for every data series was selected based on the lowest value of Akaike Information Criterion (AIC), Bayesian Information Criterion (AIC), Root Mean Square Error (RMSE), and the diagnostic tests of significance of the parameters. The normality and non-autocorrelation of the residuals of the best-fitted models were also confirmed using Jarque-Bera and Ljung-Box test, respectively. Finally, the behavior of the cancer deaths for the period 2018-2025 was forecasted using the best-selected model. The graphical comparison between historical and the forecasted data also indicate that the fitted model behaved statistically sound during and beyond the estimation period.

Keywords: Cancer, Death, Bangladesh, World, ARIMA, Forecasting.