Aggregation and Disaggregation of Information: A Holistic View

This dissertation proposes a holistic view to explore the intrinsic connections between intelligence aggregation and signal disaggregation using optimization models. We first present an algorithm that can decompose aggregated information into smaller constituents for multiple systems. This model simultaneously disaggregates integrated energy signals from over 20 houses in the U.S. into specific measurements of appliances/activities. This process provides managerial insights for energy services without installing additional hardware. Unlike some traditional approaches in the literature that require large training datasets, our disaggregation algorithm uses contextual features as inputs combined with a transfer learning approach to improve model accuracy and interpretation. The second part of this proposal develops a new optimization procedure to aggregate information from multiple sources. Our model accurately predicts a stock’s future performance by capturing inter-correlations and quantifying the reliability scores of analysts’ recommendations from financial service industry in the U.S. Given the promising experimental results, we further extend this model to obtain integrated decisions in a crowdsourced system. Considering disaggregation as the inverse process of aggregation, we advance the research on both aggregation and disaggregation of information by incorporating various types of inter-correlations in complex systems. We will further improve the effectiveness and efficiency of this framework on large-scale data from different application fields.

Many thanks to Yuyue’s dissertation committee: • Committee Chair – Hande Y Benson – Professor of Decision Sciences and MIS – Drexel University • Committee Co-Chair: Chuanren Liu – Assistant Professor of Business Analytics & Statistics – University of Tennessee • Committee Member Jinwook Lee – Assistant Professor of Decision Sciences and MIS - Drexel University • Committee Member: Muge Capan – Associate Clinical Professor of Decision Sciences and MIS – Drexel University • Committee Member: Hugo Woerdeman – Professor of Mathematics – Drexel University