Dr. Matthew Schneider, Assistant Professor at Drexel University

Title: The Effects of Data Protection on Forecasting Models

Abstract: Data protection regulations will have profound effects on the usability and quality of personal data worldwide. Consequently, the practice of forecasting will be negatively impacted due to potential alterations in these data. To measure the severity of this problem, we quantified reduction in forecast accuracy by fitting forecasting models using a variety of protected data. Results revealed that it is possible to mitigate loss of forecast accuracy by using data-driven protection methods that do not rely solely on strict legislative rules. We also show that various forms of data protection such as Aggregation can mislead a forecaster with overconfidence in the quality of their forecasts. Finally, understanding and appropriately measuring the privacy of time series data is key given the operational requirements of forecasting. A k-nearest time series swapping approach maintains the structure of the time series and performs well when balancing data privacy with forecast accuracy.

With an extensive background in data protection and privacy, Dr. Schneider will present his research on the effects on data protection on forecasting models. He will discuss upcoming data privacy legislation and innovative solutions for business to maintain the value of their analytics processes while protecting the privacy of their consumers. Joining this presentation will be Neha Khanna, an MS Business Analytics student at Drexel University who served as a research assistant on this project. Neha spent the summer with CompassRed as an intern and data scientist. She will demonstrate how the unique combination of academic research and work experience added value to her studies at Drexel University.