Theory of Time Series Analysis/Forecasting

In this course the student will learn the theory of time series analysis and forecasting.  Time series analysis is part of  artificial intelligence (AI) and is used by many companies to make predictions on sales, temperature, energy consumption, stock prices, etcetera.

Time series analysis involves looking at the time series and making judgements based on the look of the time series. The time series may need to be changed in an attempt to analyse it, and these changes could involve resampling or transforming in some fashion.

Time series forecasting involves making predictions on the time series. These predictions are based on past samples and it is important to use the correct forecasting method depending on the aoppearance of the time series, whether it has a trend or seasonality.

This course is broken down into six parts, being:-

1. Introduction

2. Time series analysis will discuss many methods used to analyse a time series, to include visualisation,smoothing, decomposition, stationarity, differencing, p-values, the Augmented Dickey Fuller test, handling missing values, the Granger causality test, autocorrelation, qqplot, lag, AIC, and RMSE.

3. Classical forecasting methods are discussed in this section, and includes the random walk, simple average, moving average, auto regression, VAR, VARMA, ARIMA, SARIMEX, SES, Holt method and Holt Winters method.

4. Machine learning and time series methods, with XGBoost and Random Forest being highlighted.

5. Facebook Prophet

6. Summary

Upon compeltion of this 45 minute course the student will be equipped with the necessary theoretical knowledge necessary to begin to code a time series analysis and forecast in the Python programming language.