I started with data collection and used various sources listed below to collect the data for natural gas consumption. Next step was to clean the data and handle any missing values. After the data processing was done, I performed exploratory data analysis to check the existing trends in the consumption pattern and the demand-supply relationship for the natural gas over the years.

Next, since I had multi-variate time series data, I had to select machine learning model that would handle such data : two important model that I used were vector ARMA and LSTM.