Unleashing Network Mining to Discover Fraudulent Businesses
The age in which we live has been called the information age. Humanity is producing data at an unprecedented pace. Modern data grows in size, complexity and ubiquity. In the talk I am going to present why network mining is a good choice for making sense of this data world. Network mining combines the search for laws and regularity of complexity science with the ability of developing scalable solutions of machine learning.
I will focus on a particular case study: the detection of fraudulent actors in a business network. Businesses are asked to report who their customer and providers were. Mismatches in these reports are a symptoms of fraudulent activities. However, without additional information it is hard to tell which side of a mismatched report is lying. I developed a technique which exploits the network of reports and iteratively updates a trustworthiness score as it uncovers more and more accurate information about the trustworthiness of the businesses. The score is a good predictor of the likelihood of a business to pay a fine, if audited.
Bio: Michele Coscia obtained his Master in Digital Humanities (2008) and his PhD in Computer Science (2012) from the University of Pisa. He then spent seven months conducting research at Northeastern University's Center for Complex Network Research, led by Albert-Laszlo Barabasi. Michele currently works as a Growth Lab Fellow at CID.
He is trained in data mining and his research is focused primarily on Complex Network analysis, particularly on multidimensional networks, i.e. networks expressing multiple different relations at the same time.
He currently works on the connection between complex networks, human mobility and knowledge flows. His research aims to understand how knowledge passes from one person to another as the result of business travels, telecommunication and commuting, and what are the repercussions of these dynamics on economic growth and development.