Forecasting Causes of Death in Northern Iraq Using Neural Network

The availability of models for predicting future events is essential for enhancing the efficiency of systems. This paper attempts to predict demographic variation by employing multi-layer perceptron network. Here we present the implementation of a system for predicting the number and causes of deaths, for a future 2-year period. The system was built using predictive models and data that is as accurate as possible under the current conditions of the northern Region of Iraq (the Autonomous Region of Kurdistan). Our predictive model is based on quarterly periods, with the intention of providing predictions on the number of deaths, classified by gender, cause of death, age at death, administrative district (governorate), and hospital where the death occurred. The data was collected from birth and death registry bureaus and forensic medicine departments for the years 2009–2020. The python programming language was used to test the designed multi-layer perceptron network with backpropagation training algorithm. With learning rate 0.01 and 500 epochs we were able to obtain good results, as the neural network was able to represent the string, and predict future values well, with a mean squared error of 0.43, and we found that number of deaths is quite stable, with a slight increase.