A deep learning model to predict volatility at earnings Announcement Dates. This model have been developed at the hackathon Artificial Intelligence & Machine Learning hosted by the House of Finance et the Quantitative Management Initiative.
The aim of the project is to set up a systematic strategy that would take advantage of the behaviour of the ( implied and realized ) volatility of individual securities around the earnings announcement dates of the companies concerned. Investments are not made directly on the underlyings but on their volatilities via simultaneous purchases of call and put (straddle). Thus the objective of this work is to predict samples for which the magnitude of the share's return, on the announcement date, is greater than expected.