Exploring and Preprocessing of Crunchbase Data for Accurate Startup Success Prediction

Startups are the backbone of the country's economy. The growth of startups is supporting the country by providing jobs and also directly generating revenue for the government. On the other hand, the failure of startups is impacting the economy's growth. Therefore, we need to explore and study the factors of startup success and failure. However, there is very fewer source of information about the startup's activity available openly, but the databases like Crunchbase are collecting and providing data about the startups for conducting study. In this paper, we are focusing on Crunchbase data for exploring the essential facts and attributes which are significantly contributing to startup success and failure. Additionally, we also proposed a machine learning model to accurately predict startup success and failure by utilizing the concluded attributes. The aim of the proposed model is to identify how the funding influences the operational status of the company. Thus, the Crunchbase dataset is considered where 66368 companies' data are used for evaluation and experiment. According to the conducted experiments we have found that venture capital financing is the main source of getting funds, 63.99% of companies are able to secure only the first round of funding, and less than 1 % of companies are able to get funding more than 10 times. The combination of the proposed preprocessing technique and learning algorithms provides highly accurate results and is able to predict 80% success and failure.