Probabilistic Prediction for a Start-Up Success Through Bayesian Networks-Based Machine Learning Approach

Abstract

Entrepreneurship propels the Indian economy forward, with the last two decades witnessing enormous growth in startup culture in India. The transformation of college campuses into hubs for entrepreneurial activities has resulted in the launch of multiple successful ventures. Over time, new startup launchpads have emerged that provide funding and growth opportunities to budding founders. Recently, an incubation program was conducted with the participation of over 2000 start-ups and the involvement of many venture capitalists and angel investors willing to invest. These startups were judged by investors on various parameters of incubation and seed funding. These parameters include scalability, employability, growth, funding, brand creation, awareness among people, marketing, etc. The selection of appropriate parameters is tedious but critical for startup growth. This study has two objectives: first, identify and establish a hierarchical ranking order of parameters that are key to success for a startup; and second, formulate a function to predict the success probability of a startup based on the correlation in these parameters. Scores will be attached to each dimension in the training data set. We create a model based on training data and perform predictive analysis using Bayesian networks to assess the probability of success for a given startup. We use Logistic regression to analyze the importance of these parameters and suggest their ranking order. This study will be advantageous to a plethora of stakeholders. Budding startups may predict roadmaps and optimize workflows. Angel investors and venture capitalists may evaluate companies they wish to invest in using the model. We modelled our study on the following parameters: Quality of the Idea, Market conditions, Strong Core Team, Market entry time, and Business model. The inferred order of the parameters is Quality of the Idea, Market entry time, Business Model, Strong Core Team, and Market conditions, respectively.