Deep neural networks, gradient-boosted trees, random forests: Statistical arbitrage on the S&P 500

Abstract

Standard-Nutzungsbedingungen:Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden.Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen.Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in AbstractIn recent years, machine learning research has gained momentum: New developments in the field of deep learning allow for multiple levels of abstraction and are starting to supersede well-known and powerful tree-based techniques mainly operating on the original feature space. All these methods can be applied to various fields, including finance. This article implements and analyses the effectiveness of deep neural networks (DNN), gradientboosted-trees (GBT), random forests (RAF), and a combination (ENS) of these methods in the context of statistical arbitrage. Each model is trained on lagged returns of all stocks in the S&P 500, after elimination of survivor bias. From 1992 to 2015, daily one-day-ahead trading signals are generated based on the probability forecast of a stock to outperform the general market. The highest k probabilities are converted into long and the lowest k probabilities into short positions, thus censoring the less certain middle part of the ranking. Empirical findings are promising. A simple ensemble consisting of one deep neural network, one gradient-boosted tree, and one random forest produces out-of-sample returns exceeding 0.45 percent per day for k = 10, prior to transaction costs. Irrespective of the fact that profits are declining in recent years, our findings pose a severe challenge to the semi-strong form of market efficiency.

Deep learning with long short-term memory networks for financial market predictions

“…Specifically, we expand on the recent work of Krauss et al (2017) on the same data sample for the sake of comparability. The authors use deep learning, random forests, gradient-boosted trees, and different ensembles as forecasting methods on all S&P 500 constituents from 1992 to 2015.…”

Random Forest Prediction of IPO Underpricing

“…The reason is that it affects the correlation among tree output and, therefore, the generalization error of the model. Krauss et al [11] show that random forests are the dominant single method for deriving trading decisions on the Standard and Poor s (S&P) 500 stock universe vis-à-vis deep neural networks and gradient-boosted trees. Other general properties that are relevant in the IPO underpricing prediction domain are its relative robustness with regard to outliers in training data and the fact that it does not overfit.…”

Pairs trading with partial cointegration

“…As in Krauss et al (2016), our choice is motivated by market efficiency, computational feasibility, and liquidity. The S&P 500 contains the leading 500 companies of the U.S. stock market, comprising approximately 80 percent of available market capitalization (S&P Dow Jones Indices, 2015).…”

“…This highly liquid subset serves as an acid test for any trading strategy, in light of significant investor scrutiny and intense analyst coverage. We proceed along the lines of Krauss et al (2016) Strategy variants: In particular, we benchmark four strategy variants against each other.…”

“…We follow Krauss et al (2016) and run a fully-fledged performance evaluation on all four trading strategies, relying on the R package PerformanceAnalytics by Peterson and Carl (2014). Most of the metrics discussed below can be found in Bacon (2008).…”