Since 1994 Zapranis has been working on neural networks and financial theory and developed numerous applications. Fluent in C/C++, Python and Matlab, has written thousands of lines of programming code and implemented advanced neural network models in C and Python. His main research themes cover the following:
Machine learning and non-parametric model estimation: development of learning procedures and neural models.
Neural model selection and adequacy testing: development of identification procedures for neural network models and diagnostics/residual analysis for model (mis-) specification.
Input variable significance testing of neural models: model sensitivity of input variables to sampling variance and parameter perturbations.
Confidence and prediction intervals of neural models: bootstrap, maximum likelihood and analytical approaches constructing confidence and prediction intervals.
Stock ranking with neural networks: neural networks as a replacement to classical statistical techniques for forecasting within the framework of the APT (Arbitrage Pricing Theory).
Tactical asset allocation with neural networks: expanding modern investment management models, that relying on the assumption that asset returns can be explained in terms of a set of factors, with neural networks which have the ability to infer complex non-linear relationships between an asset price and its determinants.
Modelling the term structure of the short rate with neural networks: non-parametric non-linear neural estimation of the reversion speed α, in the context of the Vasicek model, for deriving the term structure of the short rate.
Wavelet neural networks: mathematical aspects of wavelet neural network construction to statistical modeling and forecasting applications in areas such as finance, chaos, and classification.
Technical analysis for algorithmic pattern recognition: developing novel rule-based technical analysis pattern recognizers and implementing statistical tests for assessing their performance. Based on the algorithmic and thus unbiased pattern recognition.
Weather derivatives pricing: pricing and modeling of weather derivatives written on various underlying weather variables and strategies for effectively hedging against weather-related risk.
Dynamic time warping in financial time series: development of a dynamic time warping algorithm that can provide a similarity measure for different financial time series that might differ in length.
Asset pricing with time-varying betas: capturing time variation in a beta coefficient which is treated as a function of market return.
Herding behavior in the capital markets: investigation of the existence of herding towards market, co-skewness and co-kurtosis in developed markets and the effect of unexpected shocks in the emergence of herding.