University of Oulu

Beating the index with deep learning : a method for passive investing and systematic active investing

Saved in:
Author: Nguyen, Thi Huong Thu1
Organizations: 1University of Oulu, Oulu Business School, Department of Finance, Finance
Format: ebook
Version: published version
Access: open
Online Access: PDF Full Text (PDF, 2.6 MB)
Pages: 87
Persistent link: http://urn.fi/URN:NBN:fi:oulu-202106178452
Language: English
Published: Oulu : T. Nguyen, 2021
Publish Date: 2021-06-29
Thesis type: Master's thesis
Tutor: Conlin, Andrew
Reviewer: Sahlström, Petri
Conlin, Andrew
Description:

Abstract

In index tracking, while the full replication requires holding all the asset constituents of the index in the tracking portfolio, the sampling approach attempts to construct a tracking portfolio with a subset of assets. Thus, sampling seems to be the approach of choice when considering the flexibility and transaction costs. Two problems that need to be solved to implement the sampling approach are asset selection and asset weighting. This study proposes a framework implemented in two stages: first selecting the assets and then determining asset components’ weights. This study uses a deep autoencoder model for stock selection. The study then applies the L2 regularization technique to set up a quadratic programming problem to determine investment weights of stock components.

Since the tracking portfolio tends to underperform the market index after taking management costs into accounts, the portfolio that can generate the excess returns over the index (index beating) brings more competitive advantages to passive fund managers. Thus, the proposed framework attempts to construct a portfolio with a small number of stocks that can both follow the market trends and generate excess returns over the market index.

The framework successfully constructed a portfolio with ten stocks beating the S&P 500 index in any given 1-year period with a justifiable risk level.

see all

Subjects:
Copyright information: © Thi Huong Thu Nguyen, 2021. This publication is copyrighted. You may download, display and print it for your own personal use. Commercial use is prohibited.