ABOUT THE COURSE:Time series applications are found abundantly across different fields of research such as business, finance, climatology, environmental sciences among many others. There is a need of the hour to suitably model such datasets and try to get meaningful forecasts. A few examples in this regard are predicting the stock price of a stock, forecasting the monthly rainfall or temperature, signal processing etc. In this course we shall first introduce time series through several relevant examples, build upon the basic understanding of linear time series modelling and forecasting, while also delving deep into spectral analysis, multivariate time series, volatility modelling, and machine learning in time series. Hands-on applications and exercises will also be discussed using the R software.INTENDED AUDIENCE: Anyone pursuing BSc, MSc, MBA, MTech or PhD, with appreciation for statistics, analytics, finance etc.PREREQUISITES: Basic knowledge about statistical inference and regressionINDUSTRY SUPPORT: Financial institutions, banks, insurance groups, climatologists, economists, sales managers and other related sectors dealing with time-based observations. Eg. Commercial banks such as SBI, HDFC, ICICI etc., insurance groups such as ICICI Pru etc.,
Time Series Modelling and Forecasting with Applications in R
Indian Institute of Technology Bombay and NPTEL via Swayam
Overview
Syllabus
Week 1:Introduction to time series with examples, stationarity, non-stationarity and related conceptsWeek 2:Time series decomposition and introduction to basic time series models such as Random walk, White noise, AR, MA, ARMA etc. Introducing ACF and PACF plots and model identificationWeek 3:Tests for stationarity, expanding non-stationarity and related models such as ARIMA, SARIMA etc. Introduction to differencing and backshift operatorsWeek 4:Model identification, estimation and diagnostic checking using tests such as Augmented Dickey Fuller, Ljung and Box, etc.Week 5:Time series forecasting methods such as ARIMA, SMA smoothing, EMA smoothing, Holt Winter’s technique etc. Comparing forecasts using different metricsWeek 6:Introducing fractionally integrated processes such as ARFIMA, long memory property of ARFIMA processes, estimation of parameters etc.Week 7:Multivariate time series processes such as VAR, VARMA, moments, cross moments and stationarity, Wald representationWeek 8:Error correction models, cointegration for multivariate time series, causality analysis and causality tests, direct Granger procedure, Haugh-Pierce test, Hsiao procedureWeek 9:Fourier transformation, processes in frequency domain, spectral representation of time series, spectral densityWeek 10:Cointegration for bivariate time series, cointegration tests, Engle and Granger two-step procedure for cointegration analysis, cointegration for general multivariate processes, Johansen testWeek 11:Introduction to stochastic volatility models such as ARCH, GARCH and their extensions.Week 12:Introduction to machine learning models for time series. Anomaly detection, LSTM, Neural networks in time series
Taught by
Prof. Sudeep Bapat
