| Summary: | Despite the evolution of data science during recent years, some problems still persist when studying decision making processes. Issues such as missing data, errors, outliers, imbalance, internal correlations and the lack of unique solutions have to be properly addressed to avoid erroneous inferences. This thesis, addresses these issues in three case studies of decision making problems in the general area of credit risk management, financial investment services and financial trading. First, in the case of credit risk management, this work overcomes the problem of dealing with several scenarios that financial lenders have to face when trying to re-structure their credit portfolios. A framework is presented that allows the reduction of the solutions’ selection and in consequence improve the risk management process within these organisations. Second, within financial investment services, this thesis overcomes the challenges of profiling individual investors in the spread trading market by using ensemble data mining techniques. The application of such techniques, over this new domain, allows overcoming the complexities of profiling individual investors coming from different backgrounds in a very dynamic environment, and therefore improving the decision making process and risk management in retail brokers. Finally, within the financial trading context, by applying the appropriate controls and modelling the internal correlations in a high volume of trading data, it is revealed whether new technologies, such as smart mobiles (tablets and smart phones) and their apps, effectively help individual investors make better decisions.
|
|---|
