US Trade Balance Forecasting

In this project, we forecast the U.S. International Trade Balance in Goods and Services. International Trade Balance in Goods and Services attempts to accurately measure the trade balance of the United States, which is the difference between imports and exports measured in the millions of US dollars ($).

Data

The original publisher of the data we use is the Bureau of Economic Analysis (BEA). However, we obtained the data from the Federal Reserve Economic Data (FRED) using the FRED code BOPGSTB.

This data is present in BOPGSTB.csv.

Evaluation Criteria

For the purpose of comparing different models we use for our data, the metric we use is the Akaike Information Criterion (AIC) to choose the best forecasting model. Since we are just beginning to work out a model for the Trade Balance time series, we believe that all models we come up with are just approximations of the true model, and our goal here is to find the best approximation, rather than finding the true model. Moreover, we do not have access to out-of-sample data. These are the reasons for us to choose the AIC criterion over others like the Bayesian Information Criterion (BIC).

Repository API

The main files and directory of this repository are as follows:

• ./images: The directory containing all output images generated by the python script
• report.py: The main Python script for generating all the results
• BOPGSTB.csv: The .csv file containing the data of the time series

Model

We choose the ARIMA model for building our point and interval forecasts. More information about why we chose this and the exact parameters we use can be found in Project Report.pdf.