This project is proposed for my curiosity in exploring the influence of Brexit on the exchange rate, some parts may not be highly completed.
Firstly, it makes a rough prediction only according to the historical data of the exchange rate of GBP to EUR, by reproducing/using a specific Clockwork RNN(Jan Koutník, 2014).
Secondly, we have three features so far for every Brexit event(BE), the first feature of every BE is its inherent type, finished by Womble Bond Dickinson (UK) LLP, as: 0.Case 1.Consultation 2.Event 3.General 4.Legislation 5.Negotiation 6.Mixed(additionally added); The next feature is its strength/ability, defined as in what extent can it suddenly change the previous tendency, times how long can the later tendency last; And the last one is imported from another dataset -- the public opinion of UK citizens about Brexit, it shows who many people want to leave while other people not.
Finally, it is going to optimise/refine the rough tendency by learning the features of BEs through a multilayer perceptron, with the n input(n is the number of BE features), and m output(m is the number of how many days after that BE). We concern the differences between m values of the true exchange rate and the m values which are predicted by cwrnn as the expected values of m output, so the errors are the differences between the output values and their expected value, then adjust the weights by backpropagation those errors. We train it with all happened BEs iteratively, and we assume the features that new BE has. Input there features into the trained model, finally attain the more precise future tendency by adding every output value on each date it corresponds to.
The authorised data of exchange rates come from the online database https://fred.stlouisfed.org/categories/15 (The file 'GBP2EUR.csv' is merely an example that shows the shape of the final data we requested from the internet.)
The Brexit event data 'BREXIT.csv' comes from the website https://www.womblebonddickinson.com/uk/insights/timelines/brexit-timeline
[1] Koutnik, J., Greff, K., Gomez, F. and Schmidhuber, J., 2014. A clockwork rnn. arXiv preprint arXiv:1402.3511. [online] Available at: https://arxiv.org/abs/1402.3511
[2] Federal Reserve Bank of San Francisco, 2017. Brexit: Whither the Pound? [online] Available at: https://www.frbsf.org/economic-research/publications/economic-letter/2017/april/brexit-whither-the-pound/
