Prediction of ESG compliance using a heterogeneous information network

Abstract

Blacklists are widely used in society to avoid interactions with inappropriate entities. For example, international organizations issue sanctions lists that are used to prohibit trade with entities that are involved in illegal activities. Financial institutions keep blacklists of inappropriate firms that have financial problems or environmental issues. They create their blacklists by gathering information from various news sources to keep their portfolios profitable as well as green. In the present paper, we focus on the prediction of blacklists in the finance domain. We construct a vast heterogeneous information network that covers the necessary information surrounding each firm, which is assembled using seven professionally curated datasets and two open datasets, which results in approximately 50 million nodes and 400 million edges in total. Exploiting this vast heterogeneous information network, we propose a model that can learn to predict firms that are more likely to be added to a blacklist in the near future. Our approach is tested using the negative news blacklist data of more than 35,000 firms worldwide from January 2012 to May 2018. Comparing with the state-of-the-art methods with and without the network, we show that the predictive accuracy is substantially improved when using the heterogeneous information network. This work suggests new ways to consolidate the diffuse information contained in big data to monitor dominant firms on a global scale for better risk management, more socially responsible investment, and the surveillance of dominant firms. A B C D E P Q (b) Simplified network that illustrates the investigation.that can adjust its prediction strategy to each blacklist category accordingly. Thus, we aim to build a model that can adaptively adjust to each category. However, it is not sufficient to develop an adaptable prediction strategy for each blacklist category by using only basic information that one data vendor provides (i.e., date of addition, industry classification, and headquarters location). Thus, we construct a vast heterogeneous information network that covers the necessary information surrounding each firm by gathering information from several sources. The network is assembled using seven professionally curated datasets and two open datasets, which results in approximately 50 million nodes and 400 million edges in total. Exploiting this vast heterogeneous information network, we propose a model that can navigate through the network to predict firms that are more likely to be added to each blacklist in the near future.To motivate the heterogeneous information network approach in our setting further, we provide a specific example of how real investigators and journalists solve the problem of determining possible entities to add to the blacklist. This example is from a book written by a former member of the United Nations Panel of Experts on Sanctions against North Korea [10]. The Panel of Experts is in charge of the investigation to determine possible candidates to include in the Unit...

Effectiveness of using BIM in construction: comparative analysis and evaluation methodology

“…The options for using BIM Uses are different: from using BIM to solve one problem at one stage and ending with the integrated use of BIM Uses throughout the entire life cycle of an investment and construction project. For example, in some cases, an information model is created to calculate the scope of work and obtain estimates, and in other cases, to coordinate participants in an investment and construction project [8].…”

Insights into Search Engine Optimization using Natural Language Processing and Machine Learning

“…Label propagation is another essential target to be achieved in SEO and it becomes quite challenging in presence of heterogenous information. Study in such problem is addressed by Hisano et al [40] by storing a voluminous information in the form of a network followed by applying Jacobian iteration for learning weights. This technique also contributes towards performing better analysis.…”

Heterogeneous Ensemble for ESG Ratings Prediction

“…For example, authors tried to increase interpretability using a fuzzy expert system (FES) [60] or a fuzzy analytic network process (FANP) [62]. Arguing that the reports analysed by ESG rating agencies were too unreliable, other researchers analysed Twitter data [40] and other online sources using natural language processing (NLP) [5] or used a heterogeneous information network [23] to achieve more accurate ESG ratings. García et al [20] theorized that certain companies should receive better ESG ratings because of their financial situation.…”

“…Some partly relied on human experts [60] which exposes them to the same issues traditional ESG ratings face. Others looked at the problem from an internal perspective [62] or created a binary "bad or not" rating [23] not comparable to standard sustainability ratings. Some authors relied exclusively on textual data [5,40] whose quality is not guaranteed.…”