Comparison of predictions of behaviour of individual investors

Abstract

Abstract. Constantly evolving rational and irrational behaviour of individual investors in financial markets explains the limitations of efficient market hypothesis. Forecasting financial markets with increased volatility allows investors to retain investments when the market is ‘going crazy’. There are various ways to measure individual investor sentiments, including surveys, calculating the sentiment index and sentiment analysis of text. This study aims to compare different individual investor sentiment forecasting methods: econometric, text analysis and artificial intelligence. Trends, moving averages and Bollinger bands are standard econometric forecasting methods. They are required to have historical data strings, which allows for the assessment of prediction accuracy. Sentiment analysis from macroeconomic analyses and forex news predicts instant states of the finance market, not assessing the past. The ensemble of Evolino recurrent neural networks (EERNN) was successfully tested to predict the exchange rates. The prediction of sentiment survey data is a new investigation for the authors, and the prediction of EERNN is a distribution of expected values reflecting the probabilities of different states of market sentiments. The selected prediction method can be used to develop the trading strategy by combining it with other trading indicators. These studies can be applied to the activities of investment funds. An innovative model based on artificial intelligence and tested early in the exchange market was adopted to forecast individual investors' sentiments and create new opportunities for investors.