Machine Learning and Cross-Sectional Returns - An Empirical Analysis of Machine Learning for Return Prediction in the Norwegian Equities Market

Aug 2, 2026

Upload by Eden

This thesis evaluates tree-based machine learning models for predicting future stock returns for constituents of the Oslo Stock Exchange All Share Index (OSEAX). Random Forest and Gradient Boosted Trees are compared with Logistic Regression, alongside benchmark performance and econometric baselines. Predicted returns drive daily and monthly long-short portfolios, using a diverse monthly feature set that incorporates established capital market anomalies, while daily features rely on momentum factors and select technical indicators. Backtesting shows daily strategies do not outperform OSEAX after transaction costs, whereas monthly strategies generate consistent excess returns with a mean annual Sharpe ratio of 0.68, challenging strict efficient market views.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages82

Reading Time3 h 27 min

AccessFree to download

Introduction

Problem definition

Literature review

Thesis structure

Theoretical framework

Efficient market hypotheses

Machine learning

Machine learning and return prediction

Data

Data preprocessing

Methodology

Features

Models

Portfolio construction and trading system

Trading signals and execution

Study periods

Transaction costs

Master’s thesis NT NU Norwegian University of Science and Technology Faculty of Economics and Management NT NU Business School Jakob Netskar Machine Learning and CrossSectional Returns An empirical analysis of machine learning for return prediction in the Norwegian equities market Master’s thesis in Economics and Business Administration Supervisor: Christian Ewald May 2023 Jakob Netskar Machine Learning and Cross-Sectional Returns An empirical analysis of machine learning for return prediction in the Norwegian equities market Master’s thesis in Economics and Business Administration Supervisor: Christian Ewald May 2023 Norwegian University of Science and Technology Faculty of Economics and Management NTNU Business School Abstract This study uses tree-based machine learning models for prediction of future stock returns of the constituents ofthe Oslo Stock Exchange All Share Index (OSEAX) . Random Forest and Gradient Boosted Trees are measured against the benchmark as well as Logistic Regression, a less complex machine learning model extensively used in traditional econometric research. Long-short portfolios rebalanced both daily and monthly are constructed based on the predictions produced by the machine learning models. A diverse feature space is used for the monthly predictions, including established capital market anomalies found in the literature. The features for the daily predictions solely consist of pure momentum factors, utilizing lagged returns with varying intervals from the past trading year in addition to a few technical indicators. The result of the empirical research presents a nuanced picture of the usefulness machine learning models exhibit in return prediction. Backtesting the portfolios show that the daily portfolio is not able to outperform the OSEAX index after accounting for transaction costs, yielding a Sharpe ratio of 0.54, equal to that of the index. The monthly portfolio does however yield consistent excess returns, producing a mean annual Sharpe ratio of 0.68. This suggests that machine learning models utilizing a diverse feature set with a longer prediction horizon can capture information not incorporated in the stock price, thus challenging the views imposed by the efficient market hypotheses. Contents 1. INTRODUCTION...................................................................................................................................... 1 1.1 PROBLEM DEFINITION.................................................................................................................................. 2 1.2 LITERATURE REVIEW................................................................................................................................... 3 1.3 THESIS STRUCTURE ..................................................................................................................................... 7 2. THEORETICAL FRAMEWORK ............................................................................................................. 8 2.1 EFFICIENT MARKET HYPOTHESES ................................................................................................................ 8 2.2 MACHINE LEARNING ................................................................................................................................... 9 2.2.1 Tree-Based Methods ......................................................................................................................... 10 2.2.2 Random Forest ................................................................................................................................. 11 2.2.3 Gradient Boosted Trees .................................................................................................................... 12 2.3 MACHINE LEARNING AND RETURN PREDICTION ....................................................................................... 12 2.3.1 Challenges Applying Mac

FAQs

Which machine learning models are used to predict stock returns, and how are they compared?

The study uses Random Forest and Gradient Boosted Trees and compares them against a Logistic Regression benchmark widely used in traditional econometric research, evaluating them against benchmarks and each other.

How are the daily and monthly long-short portfolios constructed?

Long-short portfolios are rebalanced both daily and monthly based on model predictions. Daily predictions use momentum factors from lagged returns and a few technical indicators, while monthly predictions use a broader feature space including established capital market anomalies.

What do the backtesting results show after accounting for transaction costs?

The daily portfolio does not outperform the OSEAX index after transaction costs, producing a Sharpe ratio of 0.54 equal to the index. The monthly portfolio yields consistent excess returns, with a mean annual Sharpe ratio of 0.68.

Aug 2, 2026

Upload by Eden