Speaker: Mohammad Ghassemi, PhD & Tuka Alhanai, PhD
Talk Title: Spotting Survivors: Data and Methods to Predict the Fate of Startup Venture. (45-Minute Technical)
Mohammad Ghassemi, PhD from MIT
Mohammad Ghassemi is a doctoral candidate at the Massachusetts Institute of Technology. As an undergraduate, he studied Electrical Engineering, graduated as a Goldwater scholar, and was awarded the university's highest engineering distinction. Mohammad later obtained an MPhil in Information Engineering at the University of Cambridge where he was a recipient of the prestigious Gates-Cambridge Scholarship. Since arriving at MIT in 2011, he has perused research at the interface of machine learning, health, and behavioral science. He has published over 20 journal and conference papers in top artificial intelligence and medical venues including Intensive Care Medicine, Science Translational Medicine, Nature Scientific Data and The Proceedings of the Association for the Advancement of Artificial Intelligence. Mohammad’s work has been covered by venues including The Wall Street Journal, Wired, and Newsweek.
Tuka Alhanai, PhD Candidate from MIT
Tuka Alhanai is a PhD candidate in the Department of Electrical Engineering and Computer Science at MIT. Her technical expertise lies in the domains of Artificial Intelligence, Natural Language Processing, Speech and Language Processing, and User Interface Design. At MIT, Tuka is working with over 10 years of longitudinal data from the Framingham Heart Study to identify early audio biomarkers of Dementia, Alzheimer, and Depression. Tuka publishes her work in leading international Artificial Intelligence and Signal Processing conferences including AAAI and Interspeech. She is also the recipient of multiple national and international awards for her entrepreneurial leadership, including MIT’s Legatum Prize, The Pioneers of Innovation Award, and the MassChallenge. Her pioneering work has been publicized by several media outlets including The BBC, Newsweek, Wired, and TechCrunch.
We investigate how the composition of early-stage start-up teams, and the properties of their ventures, predict their nomination to a premier entrepreneurship competition, and their continued operation two years following. We collected a novel dataset of 177 ventures, comprising 374 individuals. The dataset contained the characteristics of the entrants, free-text descriptions of the ventures, and crowd assessments of venture ideas. Using sixteen descriptors of each venture, we trained several machine learning models to predict both the nomination of the teams by the competition judges and the survival of the ventures two years later. The best performing model exceeded the performance of the competition judges in predicting venture survival. Importantly, our model was well-calibrated, facilitating its use in a real-world setting. We conclude that while immense personal commitment, professional aptitude, and market volatility have major roles in the destiny of ventures, the quantifiable initial conditions of teams also carry predictive weight. Our results have implications for entrepreneurship specifically, and team building in general.
• Get free access to more talks like this at LearAI:
• Facebook: https://www.facebook.com/OPENDATASCI/
• Twitter: https://twitter.com/odsc & @odsc
• LinkedIn: https://www.linkedin.com/company/open-data-science/
• EAST Conference May 1-4: https://odsc.com/boston