|
In 2020, under the leadership of our Deputy Director for Industry Prof. Chen Kan, RMI has launched some research collaborations on industry projects, which aim at advancing the industrial applications of new research in the field. Below are the two new industry projects that are currently in the pipeline:
1. Researchers at RMI have collaborated with GIC on modeling extreme interdependencies among financial assets for risk evaluation and portfolio construction. A preliminary study of modeling the asset correlation structure using vine copula has been completed. The implementation of the vine copula captures the increased correlation at the negative tail of the asset returns. This model gives rise to a better estimate of the tail risk (measured in terms of conditional VaR) of a multi-asset portfolio.
2. RMI researchers have also finished a working paper on the text-based industry classification. This research explores the use of various word embedding schemes and clustering algorithms for industry classification. BERT, word2vec, doc2vec, and latent semantic indexing are used for word embedding, while greedy cosine-similarity, k-means, Gaussian mixture model, and deep embedding for clustering are used as clustering algorithms. The research also focuses on construction of industry classification based on the business description extracted from the profiles of the listed companies in the US and Chinese markets. The study shows that the use of LSI as the word embedding scheme together with the k-means clustering algorithm gives an industry classification that is comparable to the standard GICS classification. This indicates that a business description of a moderate length (300 words on average) contains sufficient information about companies' business for good informative industry classification.
|
|