Trading ebook Finance - Neural Prediction of Weekly Stock Market Index

It predicts the correct trend of the stock index 43

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: bull run consistently, one week in advance. It predicts the correct trend of the stock index 43 times out of the test sample of 50 points. The maximum percentage error expressed as a percentage of the desired output is only 4.044 %, while the average error is only 0.95%. Thus, the networks performs extremely well and this performance can be suitably used to time the market. Fig. 9: Training and testing data (S&P500 index) Training data 0.8 0.6 0.4 0.2 50 Testing data 0.8 100 150 Data number 200 250 0.6 0.4 0 10 20 30 Data number 40 50 12 Fig 10: Training and test data for interest rates Long rate 1 0.5 0 50 Short rate 1 150 200 250 300 250 300 250 300 48 week delay 0.5 0 50 1 Short rate 100 100 150 200 20 week delay 0.5 0 50 100 150 200 Even after detrending and normalization, which is based on training data only, the desired output from the network is higher than the values it is trained on (Fig. 9). As such, the network is able to predict an increase in the output even if it is not trained for the exact values of the ou...
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.

Ask a homework question - tutors are online