Knowledge integration in currency prediction using a transductive inference through past data sets
Ever since the general move towards a floating exchange rate has been made by many countries, researchers attempted to find ways of explaining trends and movements. This thesis attempts to explain such trends by using the Polynomial Regression Based Transductive Learning algorithm into multi series of currency parings, using multi series data in inferring to past currency trends. Firstly, the original algorithm is analyzed for modification to multi series data, and then currency data is prepared and used. Secondly, the modified algorithm is tested with the original algorithm to assess its accuracy. The results, however, proved that the modified algorithm suffers in accuracy when compared to its original counterpart. Hence, Polynomial Regression Based Transductive Learning algorithm has not produced the expected result for multi series regression for currency series.
B00749 | (wh) | Available |
No other version available