The virtue of simplicity: On machine learning models in algorithmic trading

Hansen, Kristian Bondo

doi:10.1177/2053951720926558

Cited by 33 publications

(22 citation statements)

References 61 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…One common approach to understanding how the financial wealth system has shifted has been to pay attention to the rise of ‘quants’, that is, hard scientists, mathematicians and quantitative analysts, and reflect on how they think about markets and financial problems (Derman, 2004; Hansen, 2020a, 2020b; Souleles, 2020a, n.d.; see also Lowrie, 2017). This approach looks at the way in which quants apply the techniques and tools of their own professional life (data modelling, hypothesis testing, algorithmic development and big data analysis) to the world of finance.…”

Section: John Henry Reduxmentioning

confidence: 99%

How to think about people who don’t want to be studied: Further reflections on studying up

Souleles

2021

Critique of Anthropology

View full text Add to dashboard Cite

It is now routine for anthropologists to study those who exercise power and control wealth and status in any number of societies. Implicit in anthropology’s long-standing commitment to apprehending societies in their totality, and explicit in the call to study up, paying attention to power is just one of the routine things that anthropologists do in the course of their fieldwork. That said, many theoretical and ethical norms in the discipline are calibrated to allow researchers to both know about and protect those with relatively little power who made up much of anthropology’s original topical area of interests. By contrast, studying people who exercise power entails special ethical and theoretical consideration. This article enumerates some of those considerations, and suggests that anthropologists need to have coherent theories of social action in addition to theories of social meaning. The article also suggests that some canonical disciplinary ethical norms are inappropriate for the study of the powerful for empirical and practical reasons.

show abstract

Section: John Henry Reduxmentioning

confidence: 99%

How to think about people who don’t want to be studied: Further reflections on studying up

Souleles

2021

Critique of Anthropology

View full text Add to dashboard Cite

show abstract

“…Yet, once an automated algorithm is in production, the role of the human largely changes from active decision maker to passive controller. This change is evident in cases of machine-learning model deployment for investment management and algorithmic trading purposes (Hansen 2020).…”

Section: Cultures Of Model Usementioning

confidence: 99%

“…That pragmatic idealists put a lot of trust in carefully devised and rigorously tested models does not mean that they are “model dopes,” unthinkingly accepting a model’s actions (MacKenzie and Spears 2014b, 108). It rather suggests that they are aware of their limited comprehension of the actual big data processing undertaken by their models and that the worth of their judgment and domain knowledge is measured by their ability to select and parameterize a model suitable for the problem at hand (Hansen 2020).…”

Section: Cultures Of Model Usementioning

confidence: 99%

“…Another argument for reducing human interference in the operations of machine-learning models concerns the logic of such models. Whereas ordinary rule-based algorithms follow explicit rules established by the developer, machine-learning models develop rules as they learn from processing data (Bank of England and Financial Conduct Authority 2019, 6; Hansen 2020, 2). The purpose of these models is not only to be efficient, accurate, and robust but also to discover things (problem solutions, patterns, correlations, anomalies) that were neither specified nor necessarily possible to contemplate in advance (Mullainathan and Spiess 2017, 88).…”

Section: The Displacement Of Judgment In Complex Model Usementioning

confidence: 99%

See 1 more Smart Citation

Model Talk: Calculative Cultures in Quantitative Finance

Hansen

2020

Science, Technology, & Human Values

Self Cite

View full text Add to dashboard Cite

This paper explores how calculative cultures shape perceptions of models and practices of model use in the financial industry. A calculative culture comprises a specific set of practices and norms concerning data and model use in an organizational setting. Drawing on interviews with model users (data scientists, software developers, traders, and portfolio managers) working in algorithmic securities trading, I argue that the introduction of complex machine-learning models changes the dynamics in calculative cultures, which leads to a displacement of human judgment in quantitative finance. In this paper, I distinguish between three calculative cultures: (1) an idealistic culture of undivided trust in models, (2) a pragmatic culture of skepticism toward model accuracy, and (3) a pragmatic idealist culture of early stage skepticism and implementation and production-phase idealism. Based on the empirical material, the analysis engages with examples of each of the three calculative cultures. The study contributes to the social studies of finance and science and technology studies more broadly by showing how perceptions of models shape and are shaped through model work in data-intensive, computerized finance.

show abstract

“…The real time stock trading data can be accessed through online platform (example: Yahoo Finance). Data processing methods can be used to reduce the noise in the stock data [12]. This improves the quality of the stock market prediction.…”

Section: Introductionmentioning

confidence: 99%

Application of multilayer perceptron to deep reinforcement learning for stock market trading and analysis

Sagiraju

Shashi

2021

IJEECS

View full text Add to dashboard Cite

Trading strategies to maximize profits by tracking and responding to dynamic stock market variations is a complex task. This paper proposes to use a multilayer perceptron method (a part of artificial neural networks (ANNs)), that can be used to deploy deep reinforcement strategies to learn the process of predicting and analyzing the stock market products with the aim to maximize profit making. We trained a deep reinforcement agent using the four algorithms: proximal policy optimization (PPO), deep Q-learning (DQN), deep deterministic policy gradient (DDPG) method, and advantage actor critic (A2C). The proposed system, comprising these algorithms, is tested using real time stock data of two products: Dow Jones (DJIA-index), and Qualcomm (shares). The performance of the agent linked to the individual algorithms was evaluated, compared and analyzed using Sharpe ratio, Sortino ratio, Skew and Kurtosis, thus leading to the most effective algorithm being chosen. Based on the parameter values, the algorithm that maximizes profit making for the respective financial product was determined. We also extended the same approach to study and ascertain the predictive performance of the algorithms on trading under highly volatile scenario, such as the pandemic coronavirus disease 2019 (COVID-19).

show abstract

The virtue of simplicity: On machine learning models in algorithmic trading

Cited by 33 publications

References 61 publications

How to think about people who don’t want to be studied: Further reflections on studying up

How to think about people who don’t want to be studied: Further reflections on studying up

Model Talk: Calculative Cultures in Quantitative Finance

Application of multilayer perceptron to deep reinforcement learning for stock market trading and analysis

Contact Info

Product

Resources

About