Time series data is ubiquitous, both within and outside of the field of data science: weekly initial unemployment claim, daily term structure of interest rates, tick level stock prices, weekly company sales, daily foot traffic recorded by mobile devices, and daily number of steps taken recorded by a wearable, just to name a few.

Some of the most important and commonly used data science techniques in time series forecasting are those developed in the field of machine learning and statistics. A few basic time series statistical and machine learning modeling techniques for forecasting should be included in any data scientist’s toolkit.

This presentation discusses the formulation Vector Autoregressive (VAR) Models, one of the most important class of multivariate time series statistical models, and neural network-based techniques, which has received a lot of attention in the data science community in the past few years, demonstrates how they are implemented in practice, and compares their advantages and disadvantages used in practice. Real-world applications, demonstrated using jupyter notebooks, are used throughout the lecture to illustrate these techniques. While not the focus in this presentation, exploratory time series data analysis using histogram, kernel density plot, time-series plot, scatterplot matrix, plots of autocorrelation (i.e. correlogram), plots of partial autocorrelation, and plots of cross-correlations will also be included in the demo.

This presentation is suitable for data scientists who have working knowledge of the classical linear regression model and a basic understanding of univariate time series models, such as the class of Seasonal Autoregressive Integrated Moving Average Models with Explanatory Variables.