Time series analysis is all about understanding how to leverage data collected in the past to make predictions about the future. In this course, we’ll survey many of the statistical techniques used to analyze the trend, seasonality and variance structures of time ordered data then establish the prerequisite skillset to utilize this analysis in the development of predictive forecasting models. The knowledge gained in the beginning of the course will be reinforced through a case study on using time series analysis to estimate product demand. The course will then conclude with a discussion of special topics that highlight some of the state of the art applications and techniques in time series.
1. Motivation – We’ll begin with a general discussion of data and describe under what conditions the data can be called a “time series”.
2. Data Manipulation – Booting up R, we’ll walk through loading and manipulating time series data using many of the popular R packages such as library(lubridate) and library(ts).
3. Exploratory Methods – In this section, we’ll discuss common statistical techniques to characterize the trend, seasonality and variance of a time series. Methods to visualize the data will be introduced alongside other descriptive statistics.
4. Forecasting – Using the predictive modeling capabilities of R, we’ll motivate and discuss the use of ARIMA, Holts-Winters and other automated algorithmic procedures to find the forecasting model that minimizes out of time error.
5. Case Study – To further highlight the use of time series in practice, we’ll walk through a sample case study of a real project involving estimating demand on a large set of products with varying sales velocities.
6. Special Topics – Here we’ll discuss recent developments in the academic field of time series analysis.