Data Analytics and Information Systems – Webinar archive

Tips on Transitioning from Academia to Industry

A Webinar presented by the Data Analytics and Information Systems (DAIS) Division
Feb. 11, 11 a.m. ET
Presenter: Andres Uribe-Sanchez, Ph.D.

After more than 8 years of working in academia, the transition to an industry environment was particularly difficult. In a blink of an eye, daily routines needed to be adjusted, the lens of my understanding had to change in to answer business related questions, as well as lost some freedom/time to explore/investigate novel approaches. In this seminar I would like to share some good practices I have seen and tried to mimic from managers and successful data scientists.

Download PDF | Download Video

Visualizing the Effects of Predictor Variables in Blackbox Supervised Learning Models

In many supervised learning applications, understanding and visualizing the effects of the predictor variables on the predicted response is of paramount importance. A short coming of black box supervised learning models (e.g. complex trees, neural networks, boosted trees, random forests, nearest neighbors, local kernel-weighted methods and support vector regression) in this regard is their lack of interpretability or transparency. Partial dependence plots, which are the most popular approach for visualizing the effects of the predictors with black box supervised learning models, can produce erroneous results if the predictors are strongly correlated. This talk discusses a new visualization approach, accumulated local effects (ALE) plot, which does not require this unreliable extrapolation with correlated predictors. It is substantially less computationally expensive and avoids the extrapolation problem that can render PD plots unreliable when the predictors are highly correlated.

Download PDF | Download Video

An Introduction to the Fairness in Machine Learning, Fundamental Concepts, and Real-World Examples

Machine learning algorithms have achieved dramatic progress nowadays, and are increasingly being deployed in high-stake applications, including employment, criminal justice, personalized medicine etc. Nevertheless, fairness in machine learning remains a critical problem.

Download PDF | Download video

Feature Extraction and Selection Using Topological Data Analysis

Topological data analysis (TDA) is rapidly emerging as one of the most general-purpose methods for feature extraction and selection in a variety of predictive data analytics applications. Based on the core idea of characterizing topological structures in noisy and high-dimensional data sets using their persistence information, TDA provides a robust framework to yield suitable features.

Analysis of Large Heterogeneous Repairable System: Reliability Data in Big Data Environment

In the age of Big Data, one pressing challenge facing engineers is to perform reliability analysis for a large fleet of heterogeneous repairable systems with covariates. 

New Formulation and Methodologies for Model Based Design

To date, engineering product design relies significantly on computer simulation models (e.g., finite element analysis or FEA) to predict product performance of interest given a set of design configurations. Model prediction could be problematic without referring to the corresponding test data because all models are built to approximate the real physical systems with some assumptions and simplifications.

Monitoring Service Quality in Hospital Emergency Departments

This presentation mostly focuses on building a statistical monitoring scheme for service systems that experience time varying arrivals of customers and have time varying service rates. 

In-process Quality Assurance in Additive Manufacturing

This webinar will survey the various approaches used for monitoring build quality in additive manufacturing (AM, 3D Printing) processes, including algorithms and sensing systems. 

Identification of Sleep Apnea Events Using Heterogeneous Recurrence

Obstructive sleep apnea (OSA) is a common sleep disorder that affects 24 percent of men and nine percent of women. It is caused by the collapse of upper airway during sleep, which subsequently leads to breath cessation and decrease of blood oxygen level that triggers arousals.

Shipping Cost Optimization Using IBM InBalance Cloud Solution

This webinar will focus on how IBM's cloud-based InBalance optimization solution can be utilized to minimize e-commerce shipping costs during order sourcing and scheduling process. We will discuss the usage of node specific carriers, carrier rate cards and transit days, regional carriers, node shipping capacity and order backlog and how InBalance enables Sterling OMS and other order management systems to make the best possible sourcing decisions.

Research Challenges and Opportunities in Wireless Charging Electric Vehicles

This webinar will introduce the innovative wireless charging electric vehicle which charges the battery wirelessly from the charging infrastructure installed under the road. The technology is innovative in that the vehicle is charged while it is even in motion. One example of the commercialized system is the Korea Advanced Institute of Science and Technology (KAIST) wireless charging EV shuttle, which is called Online Electric Vehicle currently operating in the KAIST campus.

Sparse Particle Filtering for Modeling Space-Time Dynamics in Distributed Sensor Networks

Wireless sensor network has emerged as a key technology for monitoring space-time dynamics of complex systems, e.g., environmental sensor network, battlefield surveillance network and body area sensor network. As a result, distributed sensing gives rise to spatially-temporally big data.

A GTFS-based Visualization Approach for Statewide Transit Network Analysis 

Assessing the current "state of health" of individual transit networks is a fundamental part of studies aimed at planning changes and/or upgrades to the transportation network serving a region. To be able to affect changes that benefit both the individual transit networks as well as the larger transportation system, organizations need to develop meaningful strategies guided by specific performance metrics. A fundamental requirement for the development of these performance metrics is the availability of accurate data regarding transit networks.

Data Visualization: Getting Started

In today’s connected world, we have the ability to gather extensive quantities of data from many different sources. Making good use of this data is limited by our ability to understand the data content, interpret relationships in the data, and take action on what we find. Data visualization is a key part of this process and when done well, takes advantage of how humans process inputs to accelerate our understanding of the data content. This webinar will discuss the general topic of data visualization, cover key principles in human processing of charts, discuss criteria for selecting visualization tools, and demonstrate the construction of visualization for a sample data set.

Collaborative Systems for Education, Innovation and Supply Networks

How can computer and information systems overcome the difficulties of locating, integrating and sharing distributed knowledge for better decisions and better understanding? The integration of Collaborative Intelligence with Collaborative Control Theory are reviewed, explained and illustrated: They can enable significant improvement in exiting Internet and HUB based services by optimizing knowledge sharing and decisions. This webinar will describe applications in networks of cultural exchange, education and training, healthcare, manufacturing and supply networks.