Here are the variables
Response variable: exchange rate return at five minute frequency
Explanatory variables:
exchange rate return at previous lags
Other exchange rates
Time around release of economic indicators (dummy)
Day of the week (dummy)
etc.
Can you suggest any other novel data mining technique to develop suitable forecasting model for the above problem?
yes, you can use SVM for this purpose, also you can try this model with MLP or RBF(Radial basis function) network, but empirical research suggest MLP (Multilayer perceptron) is best fitting approach for financial Data.
Seems like you are conducting an event study of some sort. In that case SVM and NNets do not offer much improvement over a standard ARMA-GARCH models or a simple OLS for that matter.
You should not expect that "new" estimation methods will provide superior results. Rather you should expand on your hypothesis and the specification of the data used. Data mining is in general a bad idea in finance, which will eventually lead to type I errors.
You can use Tanagara package (it is a free package) to estimate any time series using support vector machine, ANN and so on. Remember that you have to normalize the data base, divide it to training and testing data set and then Run SVM or ANN.
You can use WEKA software for implementing support vector regression.
Hallo. We managed to make MATLAB GUI to simplify calculations and simulations using three different optimizations and also different kernel
https://www.researchgate.net/project/Multiple-Kernel-Support-Vector-Regression-MKSVR
- Badges
- Science topic
- Similar topics
- Computer Science
- Data Mining