Βιβλιοθήκη ΟΠΑ - Ψηφιακό Αποθετήριο

ΠΥΞΙΔΑ Ιδρυματικό Αποθετήριο
και Ψηφιακή Βιβλιοθήκη

Όνομα χρήστη
Κωδικός πρόσβασης

Συλλογές :	Ιδρυματικό Αποθετήριο ΟΠΑ / AUEB Institutional Repository Σχολή Οικονομικών Επιστημών / School of Economics Τμήμα Οικονομικής Επιστήμης / Department of Economics Μεταπτυχιακές Εργασίες / Postgraduate dissertations

Τίτλος :	Prediction of stock market index movement using machine learning techniques

Δημιουργός :	Impraimakis, Filippos

Συντελεστής :	Kyriazidou, Ekaterini (Επιβλέπων καθηγητής) Dimelis, Sophia (Εξεταστής) Sakellaris, Plutarchos (Εξεταστής) Athens University of Economics and Business, Department of Economics (Degree granting institution)

Τύπος :	Text

Φυσική περιγραφή :	57 p.

Γλώσσα :	en

Περίληψη :	It goes without saying that the ability to predict the direction of stock/index is of paramount importance for the viability of the companies and individual investors. An accurate prediction of the sign of a stock index is an effective hedging strategy that can mitigate the risk level of companies. In essence, setting the risk is a mean that can yield a more efficient allocation of the companies' capital. In this study, eight different classification techniques were employed for the determination of the direction of the S&P 500. Two different approaches were used as inputs, first for the acquired principal components generated from PCA and second for our existing dataset. This comparison showed that Principal Component Analysis (PCA) negatively affect our results, except the KNN algorithm. Our experimental results verified the superiority of Support Vector Machines (SVM) in predicting financial time series.

Περίληψη :

It goes without saying that the ability to predict the direction of stock/index is of paramount importance for the viability of the companies and individual investors. An accurate prediction of the sign of a stock index is an effective hedging strategy that can mitigate the risk level of companies. In essence, setting the risk is a mean that can yield a more efficient allocation of the companies' capital. In this study, eight different classification techniques were employed for the determination of the direction of the S&P 500. Two different approaches were used as inputs, first for the acquired principal components generated from PCA and second for our existing dataset. This comparison showed that Principal Component Analysis (PCA) negatively affect our results, except the KNN algorithm. Our experimental results verified the superiority of Support Vector Machines (SVM) in predicting financial time series.

Λέξη κλειδί :	Principal Component Analysis (PCA) Ridge & Lasso regression Random forests Nested cross-validation Trading Machine learning Prediction of stock market index movement Financial forecasting Neural networks Support Vector Machines (SVM)

Ημερομηνία έκδοσης :	28-02-2019

Ημερομηνία κατάθεσης :	08-06-2019

Δικαιώματα χρήσης :	Free access

Άδεια χρήσης :

Αρχείο: Impraimakis_2019.pdf

Τύπος: application/pdf

Είσοδος