AI and Machine Learning have emerged as a central aspect of analytics which is applied to multiple domains. AI and Machine Learning, Pattern classifiers and natural language processing (NLP) underpin Sentiment Analysis (SA); SA is a technology that makes rapid assessment of the sentiments expressed in news releases as well as other media sources such as Twitter and blogs. This conference addresses and explains how to extract sentiment from these multiple sources of information and showcases the advances that have taken place in the field of financial innovation.
This conference builds on the findings of the six previous highly-regarded conferences on this topic. It highlights the recent developments in the application of AI and machine learning to trading strategies including automatic and algorithmic trading, quantitative fund management.
Elijah DePalma, Thomson Reuters
Machine learning algorithms analyze multilingual professional news and unstructured social media to provide meaningful investment signals for both fundamental and quantitative strategies. Furthermore, research applications of News & Social Media Analytics are expanding beyond equities to encompass Global Macro, Systemic Risk, Supply-Chain Economics and FX Markets. We demonstrate the importance of integrating behavioral insights into long-term investment decisions, and we discuss cutting-edge AI applications for short-term trading strategies.
Merve Alanyali, Data Science Lab, Warwick Business School, University of Warwick
A mammoth amount of data is being generated by our daily interactions with technological devices and online services. Compared to existing approaches, these new forms of data offer faster and cheaper measurements of human behaviour at a global scale. Examples I will showcase range from using online newspapers to quantify the relationship between financial news and the stock market, to analysing photographs shared on Flickr to track protest outbreaks around the world.
Peter Hafez, Chief Data Scientist, RavenPack
The emergence of big data in finance has shifted the alpha focus away from being faster to being smarter than the competition. Access to alternative data sources is considered a key input to such process. Peter Hafez, Chief Data Scientist, will provide an overview of RavenPack’s Big Data Analytics and the future development of the RavenPack product suite. In addition, he will present on his latest work on thematic alpha streams as well as providing an overview of general use cases of RavenPack data across various trading and investment applications.
Tilman Sayer and Xiang Yu, OptiRisk Systems
We have created an innovative and dynamic trading strategy for equities, with a particular focus on controlling downside risk. The mathematical concept behind the approach is called stochastic dominance, where investment decisions are based on distributions rather than moments. A major contribution of news sentiment is in the prediction of future distributions. Regression analysis on news sentiment and regime switching models are employed to digest market moods and account for changing market situations.
Moderator: Professor Gautam Mitra, OptiRisk Systems
Panellists: James Cantarella, Thomson Reuters; Saeed Amen, Cuemacro; Pierce Crosby, StockTwits; Anders Bally, Sentifi; Peter Hafez, RavenPack; Stephen Morse, Twitter.
Saeed Amen, Cuemacro
We discuss several unusual sentiment datasets for trading FX. We begin by discussing how we can infer market sentiment from TIM Group sell side trade recommendation data, and how we can use that to trade FX markets in a systematic manner. We also briefly present a model which uses Prattle sentiment data gleaned from central bank communications to trade FX. Later, we discuss the benefits of using open source software in finance. We present our own open source Python trading library, PyThalesians, which includes a live demo.
Edin Zajmovic, Thomson Reuters
Enza Messina, Professor in Operations Research, University of Milano-Bicocca
In this talk we addresses the challenges of sentiment analysis of microblogs. We show how combining post contents and network structure information may lead to significant improvements in the polarity classification of the sentiment both at post and at user level. We also discuss the potential of deep learning for enhancing the classification performance through a high level feature representation.
Peiran Jiao, University of Oxford
We contrast the impact of traditional news media and social media coverage on stock market volatility and trading volume. Stocks with high social media coverage in one month experience high idiosyncratic volatility of returns and trading volume in the following month. This result is consistent with the “stale news” hypothesis. Conversely, stocks with high traditional news media coverage experience low volatility and low trading volume in the following month. This result is consistent with some traders exhibiting overconfidence when interpreting signals from news media.
Anders Bally, Sentifi
In the early 90’s the majority of financial market participants used news mainly from services like Bloomberg and Reuters to inform themselves. 20 years later, they still do. During the same period, our society went through a communication paradigm shift. Today more than 2 Billion people walk around with mobile devices and communicate what they see and think on social media. These billions of voices, when structured, can generate insights which can help investors make better investment decisions. This presentation will touch on how Sentifi structures and delivers these insights, providing an information advantage for media platforms globally.
Enza Messina and Debora Nozza, University of Milano-Bicocca
Recently, deep learning approaches have obtained promising results across many different NLP applications.
Deep learning has in particular aimed at handling efficiently huge amount of texts in an unsupervised setting by capturing, in a intuitive way, the complexity of natural language.
We show how deep learning goes beyond the traditional “bag of words” representation by constructing a so-called "neural embedding" or vector space representation of each word or document. We illustrate how this representation can be exploited for sentiment analysis.
Pierce Crosby, StockTwits
The digitizing of disparate data has led to a new wave of “emerging data” in the industry, applicable specifically to existing risk and trading models. Translating these data into repeatable results is the new frontier for many asset managers and hedge funds. The discussion will draw specifically from examples of earnings crowdsourcing, satellite imagery forecasting, and social data as it applies to investor sentiment and volatility modeling. We will also cover the diversity of methods used to deliver these data sources and what is considered the new standard for asset managers when it comes to the consumption of data.