Detecting Statistical Arbitrage Opportunities Using a Combined Neural Network - GARCH Model

This paper proposes a hybrid computational intelligent system for the detection of statistical arbitrage opportunities in pairs of assets. The proposed methodology combines nonlinear neural network autoregressive models with GARCH parametrizations of volatility for describing the dynamics of the correction of relative mispricings. First results from this approach seem encouraging; further experimentation on the optimal sampling frequency, the forecasting horizon and the points of entry and exit is conducted, in order to improve the economic value when transaction costs are taken into account