An Artificial Intelligence approach to financial forecasting using improved data representation, multi-objective optimization, and text mining

This thesis presents Artificial Intelligence (AI) approaches to creating investment models. A novel data representation to optimize forecasting models created with a Support Vector Machine (SVM) and Genetic Programming. The representation is a pseudo financial factor model (PFFM). The results show that both algorithms were able to achieve superior investment returns with the aid of the PFFM.Finally, text mining techniques for analyzing annual reports, the first is based on n-gram profiles and CNG classification. The second approach combines readability scores and performance measures. Both methods and their combination outperformed the benchmark.Next is a multi-objective approach for making predictions of a market index with the aid of an Evolutionary Artificial Neural Network (EANN). The fitness function promoted EANNs that could identify behaviour in the market that predicated direction and magnitude. The results indicated that an EANN trained for multiple objectives was superior to models created using a single-objective optimization.