Quality Control and Consistency of AI Frameworks: Addressing Data Science Practice Misalignment Biases  
Author Brady McMicken


Co-Author(s) Mike Sturdevant; Andrei Shcheprov; Alan Cordell


Abstract Despite significant progress in development of machine learning methods and cloud computing, certain gaps remain between highly sophisticated artificial intelligence and analytics frameworks and real-world data driven practical business applications. This paper identifies and analyzes in depth four sources of such implementation gaps, called data science practice misalignment biases, and introduces practical recommendations to resolve them. To illustrate the methodology, a real healthcare and supply-chain use case is presented. The use case entails accurate predictions of shipment dates of prescribed eyeglasses for orders placed by optometry practices. The use case is formulated as an optimization problem. The paper also aims to create awareness among data scientists, business, and industry practitioners, and to help improve quality and consistency of AI-based and analytics-driven business decisions.


Keywords artificial intelligence (AI), machine learning (ML), AI quality control, optimization, business analytics
    Article #:  RQD28-185

Proceedings of 28th ISSAT International Conference on Reliability & Quality in Design
August 3-5, 2023