F70TS - Time Series and Machine Learning

Course leader(s):

• George Tzougas (Edinburgh, September)
• Hadj Batatia (Dubai, September)
• Eric Nimako Aidoo (Dubai, September)
• William Weimin Yoo (Malaysia, September)

Syllabus

1. Carry out an initial data analysis of time series data and be able to identify and remove simple trends and seasonalities. (1.1 1. Examples of time series, 1.2 2. Objectives of time series analysis, 1.3 3. Examples of simple time series models, 1.4 4. Stationary models using the definition of weak stationarity and the autocovariance function ACVF and the autocorrelation function ACF., 1.5 5. Estimation and elimination of simple trend and seasonality components, 1.6 6. Understanding of the need to test the estimated noise sequence)

2. Demonstrate a critical understanding of the main properties of time series models such as MA, AR, ARMA, ARIMA and random walk models. (2.1 1. Definition of a stationary process., 2.2 2. Definition and properties of the autocovariance function ACVF and the autocorrelation function ACF of a stationary process. , 2.3 3. Definition of a white noise process., 2.4 4. Definition of MA, AR, ARMA, ARIMA and random walk models, including their characteristic equation., 2.5 5. Properties of these models, including where relevant, 2.6 a Autocorrelation function ACF and its visualisation in a correlogram., 2.7 b Partial autocorrelation function PACF and its visualisation in a correlogram., 2.8 c Causality and invertibility, and their relationship to the roots of the characteristic polynomial., 2.9 d Representation as ARinfinity and MAinfinity models.)

3. Construct time series probability models from data and verify that the model fits, using a suitable computer package. (3.1 1. Selection of a ARIMA process for a set of data, using visualisations of the data and the differenced data, correlograms of the latter, and suitable information criteria e.g. Akaike's Information Criterion and/or Bayesian Information Criterion, using a suitable computer package., 3.2 2. Verification that the chosen model fits, using a suitable computer package.)

4. Demonstrate an understanding of forecasting and parameter estimation, including their statistical properties, for a time series model. (4.1 1. Minimum mean-square error MSE forecast of some time series, including calculation of point predictions from data., 4.2 2. Forecasting errors and intervals., 4.3 3. Estimation of the mean and autocovariances of ARIMA processes., 4.4 4. Estimation of the parameters of ARIMA processes, with concrete examples such as, 4.5 a. Yule-Walker estimators of an AR process., 4.6 b. Least-square estimators of an ARMA process.)

5. Demonstrate an understanding of the basic concepts of machine learning, such as classification, clustering, regression, supervised and unsupervised learning, and be able to analyse data using these concepts in an appropriate computer package. (5.1 1. Basic concepts of machine learning, such as classification, clustering, regression, supervised and unsupervised learning., 5.2 2. Analysis of data using these concepts, in an appropriate computer package.)

6. Demonstrate a critical understanding of generative and discriminative models and their application to statistical machine learning tasks. (6.1 1. Generative and discriminative models., 6.2 2. Application of generative and discriminative models to statistical machine learning tasks)

1. Introduction to Machine Learning. Training, Test and Validation sets, Cross-Validation (CV) procedure.

2. Bias vs. Variance trade-off.

3. Generative and discriminative techniques of machine learning

4. Supervised and unsupervised learning

5. Classification, clustering, and logistic regression techniques for machine learning

6. Time Series Classical decomposition

7. Stationary processes and moving average processes of time series

8. Autoregressive processes and autoregressive moving average processes of time series

9. ARIMA processes

10. Model building and Forecasting

11. Other models used in finance and insurance

Learning outcomes

By the end of the course, students should be able to do the following:

• Carry out an initial data analysis of time series data and be able to identify and remove simple trends and seasonalities
• Demonstrate a critical understanding of the main properties of time series models such as MA, AR, ARMA, ARIMA and random walk models
• Construct time series probability models from data and verify that the model fits, using a suitable computer package
• Demonstrate an understanding of forecasting and parameter estimation, including their statistical properties, for a time series model.
• Demonstrate an understanding of the basic concepts of machine learning, such as classification, clustering, regression, supervised and unsupervised learning, and be able to analyse data using these concepts in an appropriate computer package.
• Demonstrate a critical understanding of generative and discriminative models and their application to statistical machine learning tasks.
• Use appropriate machine learning packages to read and analyse data.
• Communicate the results and conclusions of a machine learning analysis in a manner that is clear, concise and appropriate for the audience.
• Describe the properties of a time series using basic analytical and graphical tools, Understand the definitions, properties and applicability of well-known time series models
• Use an appropriate computer package to process data Present results in a way which demonstrates that they have understood the technical and broader issues of time series analysis, data mining and machine learning.
• Fit time series models to practical data sets and select suitable models, Carry out simple statistical inference, in particular forecasting, based on the fitted models.
• Estimate and remove possible trend and seasonality in a time series, Analyse the residuals of a time series using stationary models
• Demonstrate the ability to learn independently, Manage time work to deadlines and prioritise workloads
• Demonstrate an understanding of the fundamental concepts in data mining and machine learning, and an understanding of the mathematics underpinning the data mining and machine learning techniques.

Further details

Curriculum explorer: Click here

SCQF Level: 10

Credits: 15