- Lecturer: Blanka Horvath
General Prerequisites:
The course assumes a good undergraduate level understanding of statistics, especially basic characteristics of key univariate distributions, statistical estimators (Least-Square, Method of Moments, Maximum Likelihood Estimators), confidence interval, hypothesis testing, normality tests and F-Tests. Students should also familiarize themselves with Python as a programming language (packages such as pandas, numpy, matplotlib, stastmodels, scipy.stats,sklearn and others).
Course Term: Michaelmas
Course Lecture Information: 16 hours of lectures in MT
Course Overview:
This course will focus on the analysis of financial data using modern statistical learning and machine learning methods. The approach will be on the application of these methods to financial problems, using Python as the programming language. Students will be expected to write reports on their findings, demonstrating understanding through their ability to explore data and apply relevant supervised learning methods, whilst accurately interpreting their results.
The course aims to cover the following topics:
- Model Diagnostic and Model Selection (AIC/BIC).
- Non-parametric Regression and splines
- Lasso/Ridge/Elastic Net
- Kernel Regression
- Transformations, weighted regression and heteroskedasticity.
- Principal Component Analysis (PCA)
- Logistic Regression
- Decision Trees and Random Forest
- Support Vector Machine
- Time Series Models: AR(p), MA(q)
- Time Series Models: ARMA(p,q) models
- ARIMA, models with trends and seasonality
- GARCH models
The course aims to cover the following topics:
- Model Diagnostic and Model Selection (AIC/BIC).
- Non-parametric Regression and splines
- Lasso/Ridge/Elastic Net
- Kernel Regression
- Transformations, weighted regression and heteroskedasticity.
- Principal Component Analysis (PCA)
- Logistic Regression
- Decision Trees and Random Forest
- Support Vector Machine
- Time Series Models: AR(p), MA(q)
- Time Series Models: ARMA(p,q) models
- ARIMA, models with trends and seasonality
- GARCH models