Video analysis of entrepreneurs’ pitches and funding prediction

Abstract

This research was conducted for the purposes of my master’s degree thesis and aims to create a machine learning model that can make funding predictions based on the data collected from entrepreneurs' funding pitches. Additionally, the project aims to unwrap and explain the main factors that induce investors trust and invest in entrepreneurs. The methodology required the data collection by viewing 12 seasons of the TV series Shark Tank, performing audio analysis and emotion classification. The data were collected in a spreadsheet file which was used to export descriptive statistics charts and to conduct the statistical analysis. The statistically significant results of this project are the following. As the amount asked, or the evaluation increase it is less likely that the team will make a deal. Additionally, having male speakers and presenting a product in fashion or beauty industry had a negative impact on the result. Teams that own a patent or don’t have depts or loans are more likely to make a deal. Presenters with previous business success are more likely to have a deal. Also, as the number of previous sales or the number of presenters or number of syllables of the pitch increases there are beneficial results to the deal outcome. According to the findings the emotions angry and sad had a small statistically significant correlation. The findings of this research reveal helpful information for entrepreneurs and provide a practical and important use of machine learning techniques and tools and the ways the above can be used to predict the success of an entrepreneur's pitch.