Summary

The availability of large quantity of cheap sensors brought forth by the so called “Internet of Things” has resulted in an explosion of the amounts of time varying data. Understanding how to mine, process and analyze such data will only to become an ever more important skill in any data scientists toolkit.

 In this lecture, we will work through the entire process of how to analyze and model time series data, how to detect and extract trend and seasonality effects and how to implement the ARIMA class of forecasting models. Both real and synthetic datasets will be used to illustrate the different kinds of models and their underlying assumptions.

Program

  • Understanding Timeseries

    • Empirical Examples

    • Trends

    • Seasons and Cycles

  • Processing Timeseries data

    • Timeseries transformations (diff, lag, sqrt, etc)

    • Resampling/fill methods

    • Bootstrapping/Jacknife

    • Autocorrelations and Partial Autocorrelation Function

    • Correlations of 2 timeseries

  • Random Walks

    • White noise

    • Drift

    • Smoothing/Rolling window

    • Fast-Fourier Transform

  • ARIMA Models

    • Auto-regressive models (AR)

    • Moving Averages (MA)

    • Fitting ARIMA models

    • Seasonal ARIMA models

Resources

Practical Time Series Analysis: Prediction with Statistics and Machine Learning
A. Nielsen (2019)

Practical Time Series Analysis
A. Pal, P. K. S. Prakash (2017)

Introduction to Time Series and Forecasting
P. J. Brockwell, R. A. Davis (2016)

Time Series Analysis: Forecasting and Control
G. E. P. Box, G. M. Jenkins, G. C. Reinsel, G. M. Ljung (2015)

Time Series: Modeling, Computation, and Inference
Raquel Prado, Mike West (2010)

References

Previous Editions: