The module aims to prepare students for careers where a good understanding of Machine Learning methods and Python programming is necessary or advantageous. Examples include: research careers in applied economics or finance, careers in data science, or careers in data analysis generally.

(LO1) Students will be able to define, explain and motivate a number of Machine Learning methods.

(LO2) Students will be able to use libraries in Python for Machine Learning and scientific research.

(LO3) Students will be able to produce Jupyter Notebook documents, mixing formatted text in Markdown with Python code.

(LO4) Students will gain a good general ability with the Python programming language.

(S1) Flexibility and adaptability
Students will need to learn in various directions for this module, using a variety of different resources for learning the underlying Machine Learning methodology and the programming, and this will require a degree of adaptability.

(S2) Problem solving
The module involves programming exercises.

(S3) Numeracy
The module involves a substantial amount of data analysis.

(S4) Commercial awareness
Commercial uses of the Machine Learning methodologies will be described in class. The Python libraries used in the module are widely used in the commercial world.

2 hour lecture x 12 weeks
1 hour seminar x 6 weeks
1 hour group learning x 6 weeks
114 hours self-directed learning

Students will need to spend time studying the machine learning methodology, along with the code examples and online documentation. Significant time will also need to be spent coding in Python, in order to gain confidence with the facilities and in order to complete the coursework assessment.

Main topics:

Least Squares review, and Subset Selection methods;

Principal Components regression;

Lasso and related methods;

Regression trees and random forests;

General principles for building predictive Machine Learning models;

Econometric inference in high dimensional models;

Deep Learning – several sessions covering algorithmic details (backpropagation and gradient descent, stochastic gradient descent, regularisation), and the use of Python libraries for deep learning;

Applications of Big Data in Economics.

Python libraries (indicative examples): numpy, pandas, scikit-learn, pytorch/keras, nltk. Programming environments: Jupyter Notebook, Spyder. Students are encouraged to install the Anaconda distribution of Python 3, which is freely available, as this comes with Jupyter Notebook, Spyder, and many of the required libraries.

Learning resources will be available on Canvas, e.g. Jupyter Notebooks containing code from lectures and lab sessions. Participants are encouraged to read the recommended textbooks for further detail about the Machine Learning methodologies and their application in Python, and references will be given in class. There are many good textbook treatments of the topics covered in the module.

The websites for the Python packages and libraries generally have introductory examples, tutorials and documentation, and participants on the module are encouraged to work through these. The R language will be introduced and used for one topic, where the package hdm will be used.