An Efficient Hybrid Machine Learning Framework for Predicting Start-Up Company Success Rates Using Structured Business Data

Abstract

Start-up firms are the powerhouse of economic growth and innovation, but they fail in their initial years in large numbers, posing great financial risk to investors and entrepreneurs. Precise forecasting of start-up success can aid decision-making and minimize uncertainty. The paper introduces an effective machine learning-based framework of start-up success/failure prediction by using structured business data gathered through a publicly accessible start-up dataset. The proposed framework involves modeling start-up performance using criteria like funding history, funding rounds, milestones, relationships, participant data, and organizational features. The tools of data preprocessing, such as normalization and class imbalance management based on the ADASYN algorithm, are implemented to improve the data quality and reliability of the model. To determine the most effective predictive model, a variety of machine learning algorithms are used and tested, including Logistic Regression, Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Naïve Bayes, and Random Forest. To evaluate the model performance, standard evaluation measures, including performance, precision, recall, F1-score, confusion matrix, and ROC curve are utilized. Experimental findings indicate that the classifier with the highest prediction accuracy (94 and above) is the Random Forest classifier, which outperforms the other machine learning models on the same data set. The most effective model is implemented as a Flask-based web application to enable users to enter start-up information and get real-time success forecasts as well as some general recommendations, ensuring practical usability