Valuation: A Dynamic Time Warping Approach to Sentiment Time Series Classification

Abstract

In the past two decades, the venture capital industry — buoyed by the rapid technological developments in software and machine learning — has rapidly expanded and gained a strong foothold in the public sphere. With this dramatic increase in exposure, different non-traditional players have entered the venture capital space. Even celebrities and superstar athletes like LeBron James, Kobe Bryant, Jay Z, and many more, have become prominent emissaries for the businesses that they have invested in.

The recent swell in venture capital investment begs the question as to whether it is possible to predict the value of a particular investment vertical or company through studying the language used about the industry — is this hyped-up company the next Google, or rather is it the next Theranos? By using dynamic time warping to analyze sentiment time series derived from news media, this thesis aims to explore whether it is possible to implement topic modeling and sentiment analysis techniques to study the growth trajectory of companies and industries and to model investment potential for all varieties of institutional investors.