Publications

The Impact of Textual Online Harassment on the Performance of Projects in Crowdfunding

23rd Pacific Asia Conference on Information Systems (PACIS 2019), 2019

ABSTRACT: In the consequence-free and anonymous online environment, online harassment has become a serious problem. In many crowdfunding platforms, there exists offensive speech on the project pages, which might force potential funders to leave the discussion and to give up investment. The effect of online harassment on project performance remains unknown. This study attempts to investigate to what extent the textual online harassment score and the project creator’s attitude towards textual online harassment might affect project performance. We constructed a Kickstarter panel dataset consisting of 388,100 projects and designed a novel framework and an algorithm BiLSTM-CNN to extract the textual online harassment score from comments, which can reach column-wise mean ROC AUC of 0.9463. This study contributes to crowdfunding and online harassment literature and provides important implications for reputation management of projects and crowdfunding platform design.

Recommended citation: HU, W. & ZHAO, J.L. (2019). The Impact of Textual Online Harassment on the Performance of Projects in Crowdfunding. 23rd Pacific Asia Conference on Information Systems (PACIS 2019)

Portfolio Optimization using Investor Sentiment

3rd International Conference on Smart Finance, 2018

ABSTRACT: Investor sentiment in financial market has been proved to have a significant cross-sectional relation with expected stock returns. Traditional mean-variance portfolio model assumes that all the investors are well-informed and rational. However, the existence of sentiment traders and their impact on market should not be neglected. In this study, we propose a new method and experimental design for portfolio optimization using investor sentiment in multiple stages. We also demonstrate the mathematical model of sentiment-adjusted portfolio to examine the impact of investor sentiment. For the estimation issue, we adapt two machine learning methods for portfolio optimization: regularization and cross-validation. In the future work, we plan to carry out an extensive empirical investigation of sentiment-adjusted portfolio against mean-variance portfolio and Online Moving Average Reversion strategy.

Recommended citation: HU, W. & ZHAO, J.L. (2018). "Portfolio Optimization using Investor Sentiment." 3rd International Conference on Smart Finance (ICSF2018)