A neural network approach for fundamental investment analysis: a case of Athens Stock Exchange

This paper explores investment profitability in an emerging European stock market using fundamental analysis enhanced by artificial neural networks. Using a set of accounting-based financial ratios from publicly available data source, we find that these ratios possess useful information in forecasting future stock returns of Athens Stock Exchange (ATHEX) constituent firms. By combining long and short rules, the neurally reinforced fundamental strategy surpasses the unconditional buy-and-hold rule in the holdout subperiod in terms of returns (total and annualized) and risk (volatility, downside volatility and drawdown) measures. Overall results remain consistent even in the presence of trading costs. Our findings suggest that stock prices in Greece do not fully incorporate financial statement information and thus inconsistent with the principle of market efficiency at the semi-strong form.