Han Wang

We propose a novel approach to investigate gender discrimination in venture capital (VC) firms by examining the gender gap in the leadership replacement of portfolio companies. Our approach addresses the selection bias stemming from fewer female entrepreneurs seeking VC funding; it does not rely on experimental designs and offers direct evidence of whether venture capitalists replace female entrepreneurs with male ones. We specifically examine portfolio companies in China that obtained funding between 1984 and 2018 and find that female leaders in portfolio companies are 3.1% more likely to be replaced after securing capital than their male counterparts, women are 7.8% less likely to lead a portfolio company after its leadership replacement than before it, and replacing a female leader with a male leader does not improve the probability of a successful VC exit. These results confirm gender discrimination in venture capitalist practices in China. Further tests reveal the role of Confucian culture in this gender bias.

With the rapid development of Internet technology and the reform of the education model, online education has been widely recognized and applied. In the process of online learning, various types of browsing behavior characteristic data such as learning engagement and attitude will be generated. These learning behaviors are closely related to academic performance. In-depth exploration of the laws contained in the data can provide teaching assistance for education administrators. In this paper, the random forest algorithm is used to determine the importance of factors for the relationship between 11 learning behavior data and students' psychological quality test data, a total of 12-dimensional feature data and grades, and extracts six factors that have a greater impact on grades. Through the research of this paper, the method of random forest is innovatively used, and it is found that the psychological factor is one of the six important factors. This paper innovatively uses BP neural network as the prediction model, takes six important factors as input, and establishes a complete method of online learning performance prediction. The research in this paper can help teachers monitor students' learning status, detect abnormal learning behaviors and problems in time, and make timely and effective teaching interventions and adjustments in advance according to the abnormal status of students found.