time_series_forecasting

In this series, we will build forecast using single and multiple time series data while looking for similarities and possible causality.

1. Single Time Series Forecasting and Stationary Tests:

Explore single time series data and build a forecast using ARIMA model.

2. Time Series Similarities:

Measure similarity between time series using various techniques.

1. Euclidean Distance
2. Cosine Similarity
3. Dynamic Time Wrapping (DWT)

Is there a shift in similarity seasonal pattern?

3. Multiple Time Series Forecasting and Granger Causality Testing:

Compare VAR and ARIMA models to forecast daily new COVID-19 cases for top 5 countries and then test for possible causality to see if including time series data from other countries significantly improves time series prediction using pairwise F-test score.

The data was taken from the Johns Hopkins University CSSE COVID-19 dataset, and stored as 'time_series_covid19_confirmed_global.csv'