Forecasting Methods for Big Time Series Data

  • Summer schools
  • Department of Statistics
  • Application code SS-ME316
  • Starting 2020
  • Short course: Closed
  • Location: Houghton Street, London

UPDATE: Due to the global COVID-19 pandemic we will no longer be offering this course in summer 2020. Please check our latest news on this situation here.

Forecasting for time series data is a fundamental task in various fields of science. Modern computing power and the large volume of data available through the internet means that the era of “Big Data” heavily influences how we approach the problem of forecasting. When facing a huge number of time series to forecast, with a potentially huge number of explanatory variables to look at, how should we approach the underlying forecasting problem? This course starts from basic time series estimation and forecasting. Through various real data examples from economics, finance, natural sciences and more, the course builds up the necessary knowledge through practical R programming so that at the end students can identify sensible approaches to forecasting a large number of time series.

The course comprises of two main parts. The first part introduces basic univariate time series analysis and forecasting. Main topics include concepts of stationarity and building of linear time series models, estimations and model selections for various time series models, forecasting equations for linear forecasting of time series models. The second part is on the analysis and forecasting of large time series models. Topics include multivariate time series models analysis and forecasting, dimension reduction of large time series, machine learning approach in fine tuning models for large time series, sufficient forecasting and procedures.

Comprehensive lecture notes with examples and R codes to experiment with are available on the course webpage. Lectures are complemented with R examples for clearer illustrations of concepts and practical data analysis procedures.

This course is designed for third year undergraduates, postgraduates, and professionals who are interested in time series forecasting, from the basics to modern “Big Data” approaches.

Session: Two
Dates: 13 July – 31 July 2020
Lecturer: Dr Clifford Lam and Dr Philip Chan


Programme details

Key facts

Level: 300 level. Read more information on levels in our FAQs

Fees:  Please see Fees and payments

Lecturers: 36 hours

Classes: 18 hours

 Assessments*: A mid-session take-home assessment (worth 25% of the overall grade) and a final exam (worth 75% of the overall grade).


An individual project that will be assigned on Tuesday of week two and will be due at the end of the course. It will be worth 50% of the overall grade.

Final exam:

A two-hour written exam at the end of the course. The exam will be worth 50% of the overall grade.

Typical credit**: 3-4 credits (US) 7.5 ECTS points (EU)

*Assessment is optional

**You will need to check with your home institution

For more information on exams and credit, read Teaching and assessment


Calculus and Linear Algebra at lower undergraduate level, and at least one semester of Probability and Statistics.

Programme structure

Specific Topics Covered:

The first part (40%) of the course gives a comprehensive self-contained introduction to time series analysis:

  • The role of autocovariance and autocorrelation in discrete time series
  • Weak and strong stationarity
  • ACVS and ACF for stationary time series
  • Simple time series models. White noise, moving average, autoregressive, and ARMA models
  • Gaussian process and stationarity
  • ACF and PACF, their estimation and their roles in order estimation for time series models
  • Estimation for time series models: MLE and least squares estimation. Moment estimators
  • Seasonal ARIMA model. Time series modelling diagnostics
  • ARCH and GARCH models for financial returns. Relation to ARMA models
  • Forecasting of linear time series models. Forecasting equations.

The second part (60%) of the course focuses on large time series data analysis and forecasting:

  • Multivariate time series models. Stationarity. Multivariate versions for ACVS and ACF
  • Forecasting multivariate time series models
  • Understanding the problems in analysing large time series. Conceptual differences between multivariate and “Big” time series. The need for regularisations
  • Vector ARX models. Machine learning approach in regularisation and gauge of performance
  • Importance of dimension reduction. Estimable high dimensional nonlinear models versus “non-estimable” ones
  • Factor modelling for large time series. Estimation and forecasting in linear factor models
  • Time series central subspace and examples
  • Forecasting with large time series predictors. Sufficient dimension reduction directions
  • Nonparametric unknown function estimation in sufficient forecasting.

Course outcomes

After successfully completing the course, students will be able to:

  • Systematically analyse time series data, including forecasting and gauging performance of different procedures
  • Criticise a particular model or procedure used for a certain data analysis, and suggest potential improvements
  • Write R codes for analysing large time series data sets.


Department of Statistics at LSE has a distinguished history. Its roots can be traced back to the appointment of Sir Arthur Lyon Bowley, an alumnus of the University of Cambridge, at LSE in 1895. He was appointed Chair in Statistics in 1919, probably the first appointment of its kind in Britain. The Department of Statistics was submitted jointly to REF 2014 with LSE's Department of Mathematics: 84% of the research outputs of the two departments were classed as either world-leading or internationally excellent in terms of originality, significance and rigour.

The department has an international reputation for development of statistical methodology that has grown from its long history of active contributions to research and teaching in statistics for the social sciences.

On this three week intensive programme, you will engage with and learn from full-time lecturers from the LSE’s statistics faculty.

Reading materials

  • Peter J. Brockwell and Richard A. Davis (2016), Introduction to Time Series and Forecasting, Springer Texts in Statistics, 3rd Edition.
  • Tsay (2014), Multivariate Time Series Analysis with R and Financial Applications, John Wiley. Hoboken, NJ.
  • Shumway, Robert H. and Stoffer, David S. (2017), Time Series Analysis and Its Applications - With R Examples, Springer Texts in Statistics, 4th Edition. (free download).

Lecture notes and other related materials will be available on the course website.

Sign up

Sign up

  • Please enter a valid email address. We will send you relevant material regarding the LSE Summer School programme.
  • Which course subject area(s) would you like to know more about?
  • Your privacy
    The details you give on this form will be stored on a secure database. LSE Summer School will use your data to send you relevant information about the School and to find out about your experiences of applying to LSE. The data on the form will also be used for monitoring purposes and to track future applications. LSE will not give or sell your details to any other third party organisation. Your data is subject to the LSE website terms and conditions and our Data Protection Policy. You can withdraw from our lists at any time by using the 'unsubscribe/manage email preferences' link that can be found in the footer of each email, or by contacting

How to Apply

Related Programmes

Optimization with Applications in Portfolio Choice

Code(s) SS-ME201

Introduction to Data Science and Machine Learning

Code(s) SS-ME314

Request a prospectus

  • Name
  • Address

Register your interest

  • Name

Speak to Admissions

Content to be supplied