Application of machine learning techniques to predict entrepreneurial firm valuation

Venture capital (VC) is the main contributor to entrepreneurial firms' funding and thus plays a crucial role in their sustainable development and rapid growth. However, early-stage VC investors often face valuation obstacles to predict firm valuation since entrepreneurial firms lack operational performance records and information asymmetry exists between them. In this paper, an integrated differential evolution algorithm and adaptive moment estimation method scheme (Adam-ENN) is proposed for early-stage VC investors to predict entrepreneurial firm valuation. Experimental results show that the proposed machine learning method outperforms the baseline methods. The feature contribution analysis and partial dependence plots were performed to open up the black box of the relationships between entrepreneurial firm valuation and its features. Results indicate that the number of VC investors in the funding syndicate is the most important feature, and VC investors' social capital also plays a significant role in the prediction model. Interestingly, the number of patents cannot convey an effective signal in entrepreneurial firm quality especially in the early-stage entrepreneurial firm valuation. Finally, this paper helps to guide entrepreneurial firms' valuation using machine learning techniques and offers deep insight into the entrepreneurship financing mechanism from the perspective of VC.