Study on Unemployment Rate in USA Using Computational and Statistical Methods

This research aimed to forecast economic indicators such as national unemployment rates in an effort to display the implementation of an ARIMA(Autoregressive Integrated Moving Average) model in time series forecasting. To prevent any noise from disturbing the economic data, linear and nonlinear smoothers were applied and made the time series analysis more efficient and accurate. Using MATLAB and Python, this paper visualized the data with time-series decompositions, allowing us to decompose time series into trend, seasonality and noise. The low order of polynomial graphs showed low accuracy on the data set when predicting closing values. The high order of polynomial graph showed high accuracy. However, the high order graph was susceptible to overfitting. There seemed to be a pattern that proved difficult to distinguish without plotting supplemental graphs. Results from the forecasting show that the prediction increases and approaches 5% and then converges to a horizontal line, indicating that this method was useful and most likely accurate when predicting the unemployment rate in the near future. Based on our data, however, it is nearly impossible to predict the unemployment rate in the far future without extensive information.