Predicting the Success of a Startup in Information Technology Through Machine Learning

Abstract

Predicting the success of a startup in information technology (SIT) is a very complex problem due to the diverse factors and uncertainty that affects it. The focus of automatic learning (ML) is promising because it presents good results for prediction issues; however, it presents a diversity of parameters, factors, and data that require consideration to improve prediction results. In this study, a systematic method is proposed to build a predictive model for SIT success, based on factors. The method consists of four processes, a hybrid model, and an inventory of 79 success factors. The method was applied to a database of 265 SITs from Australia with seven ML algorithms and three hybrid models based on the Voting strategy and the GreedyStepwise algorithm to reduce the factors. On average, precision increments in 11.69%, specificity in 3.25%, and accuracy in 21.75%; the prediction has precision of 82% and accuracy of 88%.