Tuesday, August 18, 2015
03:00 PM - 03:45 PM
The investment industry has attempted to make money on real or imagined patterns in everything from sunspots to twitter data. Services that automatically read websites and rapidly translate them into buy or sell signals have been around for some time and leading providers of financial news have been offering machine-reading packages for years.
Larger, tech-oriented trading outfits have also been using similar tools on a variety of text-based sources, in order to generate signals to automated trading models. Current tools mainly rely on recognition of keywords from predefined wordlists. There is plenty of talk on big data but nobody in the financial sector seems to have any leverage on large amounts of streaming unstructured language data, yet. However, the new range of tools on the horizon can learn meaning from context and don’t need to be trained.
Lars Hamberg, portfolio manager of global multi-asset funds at AFAM Funds and involved in big data projects. Lars has over 20 years experience and previously worked for UBS and ABN AMRO in various roles, including head of research. His academic record includes language studies, a LLM degree from Stockholm university and PhD studies at LTU. Lars has written a book on counter-party risk in OTC-derivatives trading, he has been arbiter at the SCC Arbitration Institute and after the Icelandic crisis he was responsible for restructuring and downsizing Glitnir AB. Featured as an industry influencer, Lars is often invited for public speaking about investment strategy and recently about “How Big Data Predictive Analytics will change the Investment Industry”. The focus of this presentation is the current capabilities and the performance of "neurologically plausible" approaches to machine learning, when using unstructured streaming (language) data as input for trading models.