Training by Rolling: Machine Learning and Stock Returns Forecasting

We execute a comparative analysis of machine learning models for the time-series forecasting of the sign of next-day cryptocurrency returns. We begin by compiling a proprietary dataset that encompasses a wide array of potential cryptocurrency valuation factors (price trends, liquidity, volatility, network, production, investor attention), subsequently identifying and evaluating the most significant factors. We apply eight machine learning models to the dataset, utilizing them as classifiers to predict the sign of next day price returns for the three largest cryptocurrencies by market capitalization: bitcoin, ethereum, and ripple. We show that the most significant valuation factors for cryptocurrency returns are price trend variables, seven and thirty-day reversal, to be specific. We conclude that support vector machines result in the most accurate classifications for all three cryptocurrencies. Additionally, we find that boosted models like AdaBoost and XGBoost have the poorest classification accuracy. At length, we construct a probability-based trading strategy that secures either a daily long or short position on one of the three examined cryptocurrencies. Ultimately, the strategy yields a Sharpe of 2.8 and a cumulative log return of 3.72. On average, the strategy’s log returns outperformed standalone investments in all three cryptocurrencies by a factor of 5.64, and Sharpe ratios more than threefold.

Download Full-text

A Note on Combining Machine Learning with Statistical Modeling for Financial Data Analysis

Risks ◽

10.3390/risks8020032 ◽

2020 ◽

Vol 8 (2) ◽

pp. 32 ◽

Cited By ~ 1

Author(s):

José María Sarabia ◽

Faustino Prieto ◽

Vanesa Jordá ◽

Stefan Sperlich

Keyword(s):

Machine Learning ◽

Data Analysis ◽

Stock Returns ◽

Prior Knowledge ◽

Statistical Modeling ◽

Financial Data ◽

Adaptive Inference ◽

Data Adaptive ◽

Financial Data Analysis ◽

Global And Local

This note revisits the ideas of the so-called semiparametric methods that we consider to be very useful when applying machine learning in insurance. To this aim, we first recall the main essence of semiparametrics like the mixing of global and local estimation and the combining of explicit modeling with purely data adaptive inference. Then, we discuss stepwise approaches with different ways of integrating machine learning. Furthermore, for the modeling of prior knowledge, we introduce classes of distribution families for financial data. The proposed procedures are illustrated with data on stock returns for five companies of the Spanish value-weighted index IBEX35.

Download Full-text

The Joint Cross Section of Option and Stock Returns Predictability with Big Data and Machine Learning

SSRN Electronic Journal ◽

10.2139/ssrn.3747238 ◽

2020 ◽

Author(s):

Ruslan Goyenko ◽

Chengyu Zhang

Keyword(s):

Machine Learning ◽

Big Data ◽

Stock Returns ◽

Cross Section ◽

Returns Predictability

Download Full-text

The Influence of Research Reports on Stock Returns: The Mediating Effect of Machine-Learning-Based Investor Sentiment

Discrete Dynamics in Nature and Society ◽

10.1155/2021/5049179 ◽

2021 ◽

Vol 2021 ◽

pp. 1-14

Author(s):

Xiaohong Shen ◽

Gaoshan Wang ◽

Yue Wang

Keyword(s):

Machine Learning ◽

Stock Returns ◽

Behavioral Finance ◽

Investor Sentiment ◽

Mediating Effect ◽

Machine Learning Method ◽

Mediating Role ◽

Finance Theory ◽

The Relationship ◽

The Empirical Analysis

This paper investigates whether and how the research reports issued by securities companies affect stock returns from the perspective of investor sentiment in China. By collecting research reports and investor comments from a popular Chinese investor community, i.e., East Money, we derive two indices that represent the information contained in research reports: one is the attention of research reports and the other is the average stock rating given by research reports; then we develop an investor sentiment indicator using the machine learning method. Based on behavioral finance theory, we hypothesize that research reports have a significant effect on stock returns and investor sentiment plays a mediating role in it. The empirical analysis results confirm the above hypotheses. Specifically, the average stock rating given by research reports can better predict future stock returns, and investor sentiment plays a partial mediating role in the relationship between stock rating and stock returns.

Download Full-text

Empirical asset pricing via machine learning: evidence from the European stock market

Journal of Asset Management ◽

10.1057/s41260-021-00237-x ◽

2021 ◽

Author(s):

Wolfgang Drobetz ◽

Tizian Otto

Keyword(s):

Machine Learning ◽

Stock Returns ◽

Network Architecture ◽

Risk Measures ◽

Predictive Performance ◽

Support Vector ◽

Learning Models ◽

Learning Methods ◽

Machine Learning Methods ◽

Machine Learning Models

AbstractThis paper evaluates the predictive performance of machine learning methods in forecasting European stock returns. Compared to a linear benchmark model, interactions and nonlinear effects help improve the predictive performance. But machine learning models must be adequately trained and tuned to overcome the high dimensionality problem and to avoid overfitting. Across all machine learning methods, the most important predictors are based on price trends and fundamental signals from valuation ratios. However, the models exhibit substantial variation in statistical predictive performance that translate into pronounced differences in economic profitability. The return and risk measures of long-only trading strategies indicate that machine learning models produce sizeable gains relative to our benchmark. Neural networks perform best, also after accounting for transaction costs. A classification-based portfolio formation, utilizing a support vector machine that avoids estimating stock-level expected returns, performs even better than the neural network architecture.

Download Full-text