Using Machine Learning in Statistical Models for Population Forecasting in Iraq

Abstract

Statistical studies are the primary focus of the independent planning method in any country, as the state and the private sector use population projections in planning. Another essential element is the provision of comprehensive services, which requires knowing how many people there will be in the future, and what they are divided by age, gender and characteristics, such as ethnicity. Geography, etc. Population forecasts are also widely used in the private sector for strategic planning, as population projections are considered one of the most important indicators used by its maker and population analyzes over time. In this research, the machine Learning by using neural networks, in addition to using a classical method, which is the ARIMA time series model, to predict population projections. The internal data was scanned in both methods, up to 70% training and 30% testing, and a five-year forecast for the Iraqi population. The research found that machine learning has an advantage in estimation because it has the lowest MAPE and higher R2. The population forecast of Iraq in 2028 is 47,516.3113 (per thousand).