KDD 2020 Virtual Conference

August 23rd to 27th

Mathematical Optimization: Bringing better decisions to Data Science

Gurobi Optimization will be exhibiting at the 26Th ACM SIGKDD Conference On Knowledge Discovery And Data Mining. This conference will be held virtually on August 23 – August 27. Gurobi Optimization will be active at this conference with demos and a virtual exhibit booth. This is a great opportunity for those interested in learning how mathematical optimization complements machine learning.

As a data scientist, Machine Learning can be an amazing technology for predicting likely future business outcomes. But if you want to take the appropriate actions in order to take advantage of those likely future outcomes, you’re going to need the help of Mathematical Optimization.

The Gurobi Optimizer captures the key features of your business problem in a mathematical optimization model, considers an astronomical number of possible solutions and automatically generates an optimal solution.

See how leading companies around the world use Mathematical Optimization to set efficient schedules, optimally assign resources, and save millions of dollars in industries such as logistics, transportation, energy, supply chain, IT and more.

Demo Resource Management Optimization
Presented by: Pano Santos
Time: August 24th, 25th, 26th, and 27th at 1 PM and 3 PM PDT
Location: Online - Visit the Gurobi KDD Virtual Booth to Join
Abstract: Large professional services companies employ thousands of experts to deliver a wide variety of services, making labor the industry’s highest expense. Current manual processes and tools used within labor resources management present many limitations leading to poor demand fulfillment, low labor resource utilization, high project delivery costs, and poor customer satisfaction. During this demo, we 1)walkthrough an optimization application demo that integrates machine learning and mathematical optimization technologies, and 2) address the fundamental problem of resource management, including how to match workforce resources with the right skills and capabilities, for the right job, at the right time, location, and cost.
Presentation Embedding Mathematical Optimization In Machine Learning
Presented by: Juan Orozco Guzman
Time: August 25th, 26th, and 27th at 10 AM PDT.
Location: Online - Visit the Gurobi KDD Virtual Booth to Join
Abstract: One of the greatest challenges in the deployment of predictive models is that decision makers do not trust them. Interpretable models are more likely to be accepted by decision makers since they are easy to explain, debug, and capable of generating insights from data. In this presentation, we use mixed integer programming (MIP) technology to train classification models with discrete linear weights. MIP technology avoids the use of approximations and enables flexibility and control during the optimization process. We present an alternative formulation of the Super-sparse Linear Integer Model (SLIM) presented by Ustun & Rudin in 2016, that runs much faster. We present numerical experiments that compare the performance of both formulations and show that the alternative formulation performs consistently better across all metrics. This supports the hypothesis that the training time can be decreased when using the alternative formulation.