Department of Computer, Control and Management Engineering

With Internet provision becoming ubiquitous, people have been increasingly turning to search engines, micro-blogging platforms, questions/answering forums, and online encyclopedias for news, information and research purposes.

The massive corpus of people’s digital actions can be viewed at any moment in time as a snapshot of their collective consciousness, reflecting their instantaneous interests, concerns, and intentions, and thus opening up new opportunities for a more precise and extensive quantification of real-world phenomena, including politic, economic and social events and trends. 

A special case of interest is the financial domain, where gathering information on people’s intentions before trading decisions were taken and revealing early signs of events (like stock market moves) may have paramount importance in presence of financial crises or other catastrophic events that result from a combination of actions, and affect humans worldwide.

In this talk I will present my research on tracking trading volumes and price returns of highly treaded stocks based on search queries, browsing activity on financial portals, and sentiment analysis of related news.  My work was based on the analysis of massive-scale logs provided by Yahoo. 

Results show that daily trading volumes of stocks traded in NASDAQ-100 are correlated with daily volumes of queries related to the same stocks. Web browsing on Yahoo Finance results in a higher predictive power than regular web searches, being able to anticipate stock trading volumes by two or three days. Finally,  the combined use of sentiment analysis of news and browsing activity of users of Yahoo! Finance allows to forecast intra-day and daily price changes, thanks to a wisdom-of-crowds effect that allows to exploit users’ activity to identify and weigh properly the relevant and surprising news, enhancing considerably the forecasting power of the news sentiment.

Bio. Dr. Ilaria Bordino got  her PhD in Computer Engineering in 2010 from Sapienza University of Rome and Pompeu Fabra University of Barcelona, and is currently a Researcher at UniCredit R&D where she works on designing graph-based algorithms for the analysis of financial networks and credit risk assessment, and on natural language processing tools for online reputation management and information extraction from unstructured text. Prior to joining UniCredit, Ilaria was a visiting Researcher at Max Planck Institute for Informatics in Saarbruecken, Germany (Fall 2010), where she designed a graph-based algorithm for named-entity disambiguation, and then a Research Scientist at Yahoo Labs in Barcelona (February 2011 - June 2015), where she worked on web information retrieval and big data mining, user behavior analysis, complex networks, social networks. While at Yahoo, Ilaria participated in the European Project FOC (http://www.focproject.eu), aimed at anticipating structural instabilities in global financial networks, and in the LiMoSINe Project (http://limosine-project.eu), aimed to develop a new truly semantic aggregation paradigm for search engines, enabling semantically structural access to multi-lingual online content. Ilaria has published her work at premier conferences such as SIGIR, WSDM, CIKM, WWW, EMNLP and ICDM. She also has been serving on the program committee of top tier conferences and journals in the areas of Data Mining and Information Retrieval.