AbstractsComputer Science

Stock market index prediction using artificial neural networks trained on foreign markets

by Marcus Nordberg




Institution: KTH Royal Institute of Technology
Department:
Year: 2015
Keywords: Natural Sciences; Computer and Information Science; Computer Science; Naturvetenskap; Data- och informationsvetenskap; Datavetenskap (datalogi)
Record ID: 1356776
Full text PDF: http://urn.kb.se/resolve?urn=urn:nbn:se:kth:diva-166774


Abstract

In this report, the location dependency of stock predicting articial neural networks(ANNs) is investigated. Five ANNs of the type feed forward network are trained on vedierent stock market indices (Denmark, Germany, Japan, Sweden and USA) and thencross-tested on the other markets to examine what impact this has on the predictionperformance. It is found that the ANNs perform similarly, regardless of the marketit has been trained on. The conclusion of the study is that ANNs trained on foreignmarkets perform comparably to a domesticly trained ANN. While the results appear tobe promising, more research is needed to be able to draw any denitive conclusions. ; Denna rapport undersoker hur articiella neuronnatverk (ANN) beror pa det geograskalaget. Fem ANN av typen feed forward network tranas pa fem skilda marknader (Danmark,Tyskland, Japan, Sverige och USA) och testas sedan pa samtliga marknaderna foratt undersoka vilken eekt detta har pa forutsagelsens korrekthet. Rapporten visar attde olika natverken presterar jamlikt oavsett vilken marknad de tranats pa. Slutsatsen forrapportern ar att ANN som tranats pa utlandska marknader presterar jamforbart medinhemskt tranade ANN. Samtidigt som resultaten ar lovande behovs det mer forskningfor att kunna dra slutgiltiga slutsatser.