Derivatives pricing using deep learning methods

Derivatives, such as options and futures, are complex financial
products. Their pricing can be complicated, particularly when
involving the exotics, non-linear products, or where the asset
that underlies these derivatives is illiquid. When working with
these, it is possible that no analytical closed form solutions
exist, and numerical approximation methods are used instead.
This research proposes to apply a subset of deep learning
models that use neural networks with hidden layers to price
complex derivatives, including the exotics and the illiquid. We
will train these models to estimate the prices based on big data
by examining a large amount of market and liquidity variables.
The main purpose is to propose a novel and realistic model for
derivative pricing by addressing the limitations of the Black-
Scholes standard model used extensively by academics and
the finance industry to price derivatives. We intend to consider
theoretically and empirically the influences of market
microstructure issues such as liquidity, transactions costs and
dealers' hedging behaviour, in pricing derivatives.