Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

MD67

3 - Strategically Combining Human Intelligence and Machine Intelligence in Clinical Decision Support Systems Long Xia, Virginia Polytechnic Institute and State University, 880 W. Campus Drive, 2069 Pamplin Hall, Blacksburg, VA, 24061, United States Healthcare is a domain in which experts’ knowledge plays significant roles in a variety of tasks. This is one of reasons why solely using deep learning models often could not achieve satisfactory results. We adopted the crowd wisdom concept and proposed a framework to strategically combine human intelligence and machine intelligence in the context of clinical decision support systems. Our experimental evaluations demonstrate that our approach can make a significant performance improvement compared with existing approaches. n MD67 West Bldg 105B Causal Inference and Machine Learning Sponsored: Information Systems Sponsored Session Chair: Wei Chen, University of Arizona 1 - Deep Learning for Overweight Prediction on Twitter Luwen (Vivian) Huangfu, University of Arizona, Tucson, AZ, United States Overweight is epidemic in the United States and elsewhere in the world, which is serious and costly nowadays. However, it’s humans’ nature to store energy in case of famine, causing overweight outbreaks on a large scale due to the increased availability and consumption of high-calorie food. We are motivated to study overweight based on the language of food on Twitter using deep learning. Our result reached a good performance, with remarkable improvement when comparing with the benchmark, revealing that deep learning can be utilized to monitor overweight issues from Twitter. 2 - An Exact and Robust Conformal Inference Method for Counterfactual and Synthetic Controls We introduce new inference methods for counterfactual and synthetic control methods for evaluating policy effects. Our inference methods work in conjunction with many modern and classical methods for estimating the counterfactual mean outcome in the absence of a policy intervention. Specifically, our methods work together with the difference-in-difference, canonical synthetic control, constrained and penalized regression methods for synthetic control, factor/matrix completion models for panel data, interactive fixed effects panel models, time series models, as well as fused time series panel data models. 3 - Completing the Online-offline Circle at the Last-mile: A Large Randomized Field Experiment Tianshu Sun, University of Southern California, 3670 Trousdale Parkway, Bridge Hall, BRI 310B, Los Angeles, CA, 90089, United States, Rongqing Han, Leon Zhu, Lixia Wu Pickups stations have the promise to reduce the logistic cost of last-mile delivery in E-commerce. At the same time, they also serve as an ideal touchpoint to connect customers and brands in the physical world and can bring benefits to the platform. In this study, we examine whether E-commerce platforms can take advantage of pick-up stations and complete the online-offline circle at the last mile. We design a large experiment with Alibaba and examine 1) Whether platforms can use online intervention to elicit customers to utilize offline pickup service (Online-to-Offline) and 2) Whether offline interaction with products can encourage customers to engage with brands and platform (Offline-to-Online). 4 - Nowcasting the Local Economy: Using Yelp Data to Measure Economic Activity Hyunjin Kim, Harvard Business School, 84 HBS Mail Center, MA, Boston, MA, 02163, United States, Edward L. Glaeser, Michael Luca Can new data sources from online platforms help to measure local economic activity? In this paper, we present evidence that Yelp data can complement government surveys by measuring economic activity in close to real-time, at a granular level, and across geographic scale. Changes in the number of businesses reviewed on Yelp can generate an algorithm that explains 29.2 percent of the residual variance in changes in the number of establishments in County Business Patterns (CBP), after accounting for lagged CBP data. The algorithm is more accurate for denser, wealthier, and more educated ZIP codes. Yinchu Zhu, University of Oregon, Eugene, OR, 97401, United States, Kaspar Wuthrich, Victor Chernozhukov

n MD68 West Bldg 105C Joint Session QSR/Practice Curated: High Dimensional Data Analysis and its Applications in System Informatics and Control Sponsored: Quality, Statistics and Reliability Sponsored Session Chair: Mostafa Reisi Gahrooei, Georgia Tech, 755 Ferst Dr, Atlanta, GA, 30318, United States Co-Chair: Kamran Paynabar, Georgia Tech, Atlanta, GA, 30332, United States 1 - Image Based Online Defect Detection and Closed Loop Quality Control for Additive Manufacturing Processes Chenang Liu, Virginia Tech, VA, United States, Zhenyu Kong A major challenge in additive manufacturing (AM) is how to ensure product quality and consistency by eliminating defects. To address this challenge, this work develops an image-based closed-loop quality control system for the fused filament fabrication (FFF) process. This system consists of a real-time image acquisition device, a high accuracy image-based classification algorithm to monitor the status of surface quality, and a PID-based controller for defects mitigation. The case studies based on actual AM experiments demonstrate the effectiveness of the proposed method. 2 - Multiple Tensor-on-tensor Regression Mostafa Reisi Gahrooei, Georgia Tech, Atlanta, GA, 30318, United States In recent years, measurement or collection of heterogeneous sets of data such as those containing scalars, waveform signals, images, and even structured point clouds, has become more common. This work addresses the problem of estimating a process output, measured by a scalar, curve, image, or structured point cloud by a set of heterogeneous process variables such as scalar process setting, profile sensor readings, and images. 3 - An Efficient Monitoring of Variance in High-dimensional Process Sangahn Kim, Rutgers University, 96 Frelinghuysen Road, Room 201, Piscataway, NJ, 08854, United States, Galal M. Abdella, Jinho Kim, Khalifa M. Al-Khalifa, Myong Kee Jeong, Abdelmagid S. Hammuda In high-dimensional processes, monitoring process variability is considerably difficult due to a large number of variables and the limited number of samples. Monitoring changes in the covariance matrix is often used for monitoring process variability under the assumption that only a few elements in the covariance matrix are changed simultaneously. In this presentation, we propose a control chart based on an adaptive LASSO-thresholding for monitoring process variability and illustrate the advantages of the proposed method through simulated and real data from both semiconductor industry and high-dimensional milling process. 4 - Learning Inter-layer Bonding Effects in 3D Printing through a Convolution Formulation Yuanxiang Wang, University of Southern California (USC), CA, United States, Qiang Huang Geometric accuracy control is critical for additive manufacturing (AM) or three- dimensional (3D) printing. The prediction and control of out-of-plane deviation of 3D printed products needs the understanding of inter-layer bonding effects. This work presents a convolution formulation for the inter-layer interactions to predict out-of-plane shape deviations. Experimental investigation using stereolithography process validates the proposed model.

238