Bayesian Changepoint Detection
- MSci dissertation, performing Bayesian analysis of a coal mining disasters data set.
- Employing inference approaches (such as Gibbs sampling, Metropolis-Hastings algorithm, sequential importance sampling) to calculate the posteriors of changepoint locations and corresponding disaster rates using a semi-conjugate prior.
- Sampling from the posterior predictive distribution to see how well the model explains my observations and implement marginal likelihood evaluation procedures to decide on the number of changepoints most suitable to explain data.
Machine Learning in Simulations
- Conducted independent research as an intern at Lancaster University on the STOR-i summer internship.
- Focused on machine learning in simulations, more specifically, simulation analytics. Researching how and why the performance of a system varies through time.
- Predicted whether simulated projects would finish early or late using multiple K-fold cross validation logistic regression on chronologically segmented systems.
- Analysed the classification’s output in order to understand how the importance of each variables in predicting the outcome of the project changes throughout time.