This preview shows page 1. Sign up to view the full content.
Unformatted text preview: 86
Fig. 12 on the shows the index values in the training and testing region for this second
case, where the training data starts from the 500th data point and is 200 data points long
while the test data extends over 75 weeks immediately following the training data and
includes the big crash. Fig. 13 shows the interest rates for the training and test regions
combined. In this case, the delayed long term interest rate is falling in the training phase
when the S&P 500 index has a rising trend, while during the testing phase it is rising very
fast, when the index is falling. The short-term rate delayed by 20 weeks is also increasing,
albeit rather slowly, in the test period. Thus, in this test case, it is expected that the effect
of rising interest rates will predict the crash. There is not much sustained momentum in the
training data of the index, since it is rising and falling or in other words, the change in
index is oscillating between positive and negative values. The network having 7 neurons in
the hidden layer is trained for 900 iterations, with a constant step...
View Full Document
This note was uploaded on 07/20/2012 for the course ECON 203 taught by Professor Girishdev during the Spring '12 term at Indian Institute of Technology, Kharagpur.
- Spring '12