چكيده به لاتين
Economic variables forecasting is one of the most important matters in governments’ decision making and for private businesses. Central bank, financial institutions, and many big companies use different economic forecasting, e.g. interest rate, unemployment rate, inflation, exchange rates, etc., when they take action or determine new policies. These forecasts play an important role when governments want to present yearly budget bills.
Unfortunately, these kind of parameters are affected by a very vast and statistical set of variables as a complicated system, so forecasting would not be an easy task. This is why old-fashioned ways of forecasting are based on qualitative parameters and surveys.
In the problem of forecasting time-series data, we have different parameters in a course of specific time frame, and the goal is to predict the future value of a target variable as closely as possible. There are many old algorithms defined for this problem from which we can mention Regression, or Bayesian Forecasting, but Machine Learning has become a popular approach in recent years.
Using old models, state-of-the-art machine learning algorithms including Random Forest, LSTM, GRU and XGBoost (a specific approach in Gradient Boosting), and an innovative algorithm based on Model Stacking Ensemble, we will introduce the best model presented and we will make a thorough comparison of different models by comparing their different error metrics which have been used in the most recent and bold researches published in Time-Series Forecasting area.
