Bankruptcy prediction has been a popular and challenging research area for decades. Most prediction models are built using traditional data such as nancial gures, stock market data and rm specic variables. We complement such * Julie and Marija contributed equally.ble model is built that combines the relational model's output scores with the structured data. We nd that this ensemble model outperforms the base model when detecting the riskiest rms, especially when predicting two-years ahead.
Many real-world large datasets correspond to bipartite graph data settings-think for example of users rating movies or people visiting locations. Although there has been some prior work on data analysis with such bigraphs, no general network-oriented methodology has been proposed yet to perform node classification. In this paper we propose a threestage classification framework that effectively deals with the typical very large size of such datasets. The stages are: (1) top node weighting, (2) projection to a weighted unigraph, and (3) application of a relational classifier. This paper has two major contributions. Firstly, this general framework allows us to explore the design space, by applying different choices at the three stages, introducing new alternatives and mixing-and-matching to create new techniques. We present an empirical study of the predictive and run-time performances for different combinations of functions in the three stages over a large collection of bipartite datasets with sizes of up to 20 million × 30 million nodes. Secondly, thinking of classification on bigraph data in terms of the three-stage framework opens up the design space of possible solutions, where existing and novel functions can be mixed and matched, and tailored to the problem at hand. Indeed, in this work a novel, fast, accurate and comprehensible method emerges, called the SW-transformation, as one of the best-performing combinations in the empirical study.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.