Tid: Tisdagen 20 maj 1997 kl 1600-1700 (Obs! Dagen och tiden)
Plats : Seminarierummet 3733, Institutionen för matematik, KTH, Lindstedts väg 25, plan 7. Karta!
Föredragshållare: Nils Hast och Robert Thorén
Titel: Multipel regression som verktyg för prediktion av svenska räntan i realtid - modellbygge och analys av finansiella faktorer på historiska data. (Examensarbete)
Sammanfattning: Statistical correlation and multiple regression techniques were used on a database consisting of about 100 financial variables, such as interest rates, currency exchange rates, and stock market indices, for predicting the over-night change of a 10 year Swedish Government Bond. The predictions are based on real-time values of explaining factors, and gives real-time predictions. A 66 percent accuracy in the direction of the change (plus or minus) was achieved using a 3-factor regression model on a 22-work day interval. The standard deviation of the difference between the actual value and the predicted value was 0.11. In reference the average value of the 10 year Swedish Government Bond was 7.67.