Coursework description

The intention of the module is for you to follow along with each of the topics, testing all the methods and ideas on your own portfolio of at least half a dozen time series. The coursework mostly stems from this activity. Make sure that two of the time series in your portfolio includes high-frequency financial data of your choice from the available high-frequency data sets on the module page on VLE. Each high-frequency data set contains minute-by-minute data for 3 years. You may not consider the full 3-year range of the high-frequency time series for the analysis; instead, you may select part of the range for your analysis.

Part 1

At the end of each of these first five topics, you will find a section providing you with a brief description of what you should be doing for the portfolio at that stage. You should submit an account of your engagement with each of those first five topics. The account should include:

1. An overview of what you have done in that topic.

2. Results in the form of graphs, error estimates, or whatever else is appropriate for that topic.

3. A copy of the software you have implemented to conduct your analyses.

4. A brief account of what you have discovered about each dataset/method you have explored.

5. An overall evaluation of how well the algorithms work and a report of the insights you have gathered.

Part 2

Further, you need to perform ACF, PACF analysis and AR(q) predictions at two different frequencies of the selected high-frequency data. For example, you may extract data at intervals of 5 minutes, 10 minutes, and conduct the analysis. Utilise Monte Carlo methods in conjunction with AR(q) modelling (with a selectable q value) to obtain a range of 25 potential future scenarios for both 5-minute and 10-minute data intervals. Employ kernel density functions to estimate probability distributions for these future scenarios. Evaluate and compare the forecasting performance of the model at both frequency intervals.

Part 3

The last part of the coursework is more open-ended. The idea is that you should try something that we have not had time to do during the lectures. Just so you know, we think the topics that will be covered will be STL decomposition, ARIMA, Vector Auto Regression, ARCH, and possibly GARCH. Some possibilities for things not covered are deep learning, recurrent neural networks, and technical trading analyses.