Predicting the success probability of startups using ensemble machine learning methods to select an acquisition strategy for successful startups.
Predicting the success probability of startups using ensemble machine learning methods to select an acquisition strategy for successful startups.
reza najafi shiraye1 seied alireza razavi2 Goodarz Saadati Moghadam3
1) Master s student in Information Technology – E-Commerce, Islamic Revolution Comprehensive University.
2) Master s student in Information Technology – E-Commerce, Islamic Revolution Comprehensive University.
3) Assistant Professor, Department of Information Technology Engineering – E-Commerce, Islamic Revolution Comprehensive University.
IOI Code
:
ABDE-ABKFGE
Publication :
6th International Conference on Advanced Research in Management And Humanities(huconf.com)
Abstract :
Investing in new businesses, especially startups, is one of the riskiest decisions in investment management; despite numerous studies in this field, selecting a method that can largely resolve this challenge or sufficiently reduce its risk is very difficult and requires careful study and investigation. In this paper, to present an AI-based model for predicting the success probability of startups, data from 923 startups available on the Kaggle website have been used. In the proposed algorithm, the data are first cleaned, and then feature engineering is applied to them. Subsequently, after removing unnecessary features, the models are compared using three base machine learning models—decision tree, random forest, and logistic regression—and three ensemble architectures—soft voting, hard voting, and stacking—and are evaluated based on eleven performance evaluation metrics. The results show that the soft voting ensemble architecture ranks first with an overall score of 0.7921 and a significant margin over the other architectures. Soft voting achieved the best scores in accuracy, F1-score, and correlation coefficient compared to its competitors. Based on the prediction results of the models among the top 10 startups, almost all models achieved similar accuracies. Examination of the models indicates that startups with a higher probability of success operate in digital domains such as web, mobile, and software.
Keywords : startup ensemble learning success prediction soft voting hard voting investment selection