Staged link prediction in bipartite investment networks based on pseudo-edge generation

Investors, including venture capitalists, invest in companies in multiple stages depending on the scale and timing. A second-stage investment involves investing a large amount in a few valuable companies; it requires careful consideration and entails more factors to be considered compared to an early-stage investment. One key consideration is the information regarding the investor who provided the initial investment to the company. Therefore, numerous studies have focused on predicting the possibility of second-stage investment based on information from the early stage of investment. However, these staged link prediction methods have limitations in their prediction accuracy when there are missing data in the early-stage investment. For instance, if the information regarding the initial investment remains undisclosed due to confidentiality reasons, the accuracy of second-stage investment prediction decreases. Therefore, in this study, an advanced method is proposed to generate pseudo-edges to predict investment outcomes that are not disclosed despite the company receiving investments. Accurate investment predictions are performed through staged link prediction. To evaluate the effectiveness of the proposed method, experiments were performed using various prediction models, and the proposed method achieved the highest prediction accuracy of 88.98%.