Evaluation and success rate prediction of innovation and entrepreneurship projects using a group recommendation algorithm

With the rapid increase in the number of entrepreneurial projects, how to scientifically evaluate the potential of projects and predict their success rates has become an important challenge for investment decision-making. This study constructs a framework for evaluating and predicting the success rate of innovation and entrepreneurship projects using the group recommendation algorithm and machine learning methods. The study is based on the data of enterprises in the IT Juzi startup company database. After data cleaning, feature engineering, and core feature mapping, a multi-dimensional feature set is formed. The group recommendation part uses a collaborative filtering (CF) approach based on weighted aggregation to rank projects for investment institutions or startup mentors, achieving priority recommendation of potential projects. The success rate prediction part is based on the XGBoost binary classification model to predict the survival and deregistration status of enterprises, and the model performance is verified through multiple indicators. The results show that the proposed group recommendation model is superior to the comparison methods in six indicators, showing strong recommendation performance. The accuracy rate reaches 0.874, higher than that of CF at 0.812 and neural CF at 0.846. Feature importance analysis also reveals the impact of key factors such as financing scale, funding rounds, industry category, and team size on recommendation and prediction. The research results provide a data-driven reference for investors to scientifically identify potential projects and optimize investment decisions in group recommendation contexts, and also provide an expandable analytical framework for relevant academic research.