Shiyang Lai

• Proposed a new type of signal processing task - conversation recognition, and designed a multi-pipeline architecture which combines various neural networks (i.e., PyanNet, XVectors, and Wave2Vec model with CTC loss) and machine learning algorithms (i.e., agglomerative clustering and mutual information) together.
• Implemented the proposed architecture to extract information about Chicago police communications to support a wide range of social science-related topics that are in the interests of the project.
• Presented the interim results of the architecture development in the 2022 IC2S2 conference and won the recognition of a large number of peer experts.

Research project led by Prof. Zhiping Fan

• Explored the economic value of social function in rental platforms and revealed the recommendation mechanism of a novel e-WOM format, travel story.
• Identified the recommendation information spillover effect and its attenuation in Airbnb.
• Investigated the survival status of shared and commercial listings in Beijing and the determinants of their survival.

Research project led by Prof. Zhuchao Yu

• Integrated wavelet decomposition and time series decomposition components on the top of a multivariate LSTMs model to predict energy consumption of four Chinese metallurgy enterprises.

Other research projects

• Designed the Susceptible-Exposed-Asymptomatic-Infectious-Recovered (SEAIR) model to infer the incubation period distribution of COVID-19 of eight countries.
• Proposed a hybrid detrended deconvolution foreign exchange network construction method (DDFEN).