The INFORMS Journal on Data Science

✍️Summary: In this work, the authors propose a kernel-based iterative regression model for enhancing time series forecasting accuracy by integrating data from multiple spatial locations. #regression #timeseriesforecasting #INFORMS #datascience

🔍 Summary (1/2): Multisensor data that track system operating behaviors are widely available in engineering systems. Yet, the possible presence of sensors whose data contain irrelevant information poses a challenge. The authors propose a functional data clustering method that

🖇Link: doi.org/10.1287/ijds.2021.00… 📝Summary (1/2): Increasingly frequent wildfires have a significant impact on solar energy production as the atmospheric aerosols generated by wildfires diminish the incoming solar radiation.

📝 Summary (1/2): Cost-sensitive classification is critical in applications where misclassification errors widely vary in cost. However, overparameterization poses fundamental challenges to the cost-sensitive modeling of deep neural networks (DNNs). To address this challenge, ...

📷 Summary: This work leverages a computational conjecturing framework to produce nonlinear bounds for continuous features and boolean expressions for categorical features based on input data. Their method recovers known patterns in data that no previous method could find.

(2/3) The authors introduce three tree-based density estimation methods for binary/categorical data. Its models are sparse, and users can specify the desired number of leaves, branches, or rules with a prior.

(2/3) "This work introduces a modeling approach, inspired by backward elimination and incorporating a cost hierarchy to minimize information loss. The cost of each condition is modeled as a function of the number of other, more expensive chronic conditions an individual has."

(1/2) This study introduces an interpretable hierarchical deep learning model for the detection of Alzheimer's disease. The model utilizes transcripts of patient interviews, employing a novel hierarchical attention mechanism to capture temporal dependencies in longitudinal data.

(1/2) This paper proposes a novel diversity subsampling algorithm suitable for large real-world data. It enjoys superior performance and is far faster than existing algorithms. It is more efficient to select a diverse subsample from a large data set by requiring the subsample to

(1/3) Credit ratings are traditionally generated from models that use financial statement and market data, which are tabular. Using machine learning, this work constructs a network of firms using U.S. SEC filings to enhance the traditional tabular data set with a corporate graph.

"(2/3) US financial crisis. We model and study how multiple financial institutions form a supply chain to create these securities. We show communities of financial institutions along the supply chain are associated with the generation of a prospectus and a group of securities."

(2/2) "..each associated with a vector of non-sensitive attribute values and a categorical sensitive attribute (e.g., gender, race, etc.), our goal is to find a clustering of the items that minimizes the loss function while imposing fairness measured by Rényi correlation."

(2/3) "The authors propose a novel registration-free solution to the post-Statistical Process Control part defect localization problem. The approach leverages a spectral decomposition of the Laplace–Beltrami operator to create a functional map between CAD and measured manifolds,

"A supervised dimension reduction model was proposed for the prognostics of applications with incomplete imaging data. The proposed supervised dimension reduction-based prognostic model outperforms unsupervised dimension reduction-based models."