Harnessing The Power of Random Forest in Predicting Startup Partnership Success

Startups are an important element in innovation and economic growth. However, startup failure is very high, so investors, governments, and startups need to predict startup success. This research develops a startup success prediction model based on random forest with multi-layer decision analytics attributes, partnership longevity predictions, importance feature ranking, cross-industry startup evaluations, and ensemble learning benefits. This research uses secondary data from Crunchbase, AngelList, and VentureBeat. The data is processed with data cleaning, feature engineering, and feature selection. The data is divided into training data and testing data with a ratio of 80:20. The training data trains a random forest model with sci-kit-learn in Python. The model is evaluated by accuracy, precision, recall, F1-score, and ROC-AUC. The testing data tested the random forest model and compared it with other models. This study also used SmartPLS to test hypotheses in a variance-based SEM model with PLS path modelling. This study used 370 respondent data for SmartPLS with SmartPLS software. The results show that the random forest model is better than other models in predicting startup success. The model provides information about important features in predicting startup success. SmartPLS results show that all hypotheses are accepted with statistical significance. The model has high explanatory power and predictive power.