| Summary: | 碩士 === 東吳大學 === 經濟學系 === 92 === With rapid develop liberalization along with financial market . The financial market is getting liberalization transferring security and internationalization. The built-up risk management becomes each banking institution''s issue chief respectively for financial institution is as important as before. This article is trying to use a Study of Artificial Intelligence (Genetic Algorithm 、Fuzzy Logic and Back Propagation Network) to Risk Management. Besides Combination with traditional VaR management theory and application discuss with Monte-Carlo simulation analyze. We also had discussion on example to make a description the result of before and after GANNs risk managements. Back Propagation Network (BPN) is a most popular Artificial Intelligence Neural Networks (ANNs) system. That is data analysis tool that have been developed to enhance some of the limitations of traditional statistic methods. A distinguishing feature can be expressed as follows: (1) BPNN do not assume that data are normal-distribution. (2) BPNN do not assume that there is a linear relationship between predictor variables and the dependent variable or outcome. (3)According to computer science and technology great progress, BPNN is provided with high speed calculate (4)BPNN have memory recall and learning ability. Generic algorithm is will problematic unfasten consider as genic chromosome, and as nature survival of the
fittest in natural selection and chromosome cross-over, replication, and change suddenly operation machine-made establish fitting melt neural net framework behind most, as mold paste kind of neural-net means again, by historical data''s study seek fitting weight most, it fitting unfasten will can at mold draw up nature competition''s environment in seek most. By means of by mold paste neural net and generic algorithm''s combination, make us can exert finance data go to calculate and forecast come out optimum investment combination.By association with BPN and GA, we can use data to calculate and to predict best portfolio. This research risk worth analysis model can divide into two kinds of types traditional cover the Monte-Carlo simulation and AI -VaR analysis. By means of tradition VAR calculate, combine generic algorithm and BPN evaluation VAR, compare tradition risk management means and GABPN(generic algorithm back-propagation neural network) VaR''s forecast performance, and proceed investment combination performance evaluation. Demonstration in the end discover, no matter select what kind of AI style simulation VAR, it forecasts performance superior to tradition Monte-Carlo simulation draw up law average, combination performance analytical method GABPN model is better be investing also.
|
|---|
