Explanation Interfaces for Sales Forecasting

Item Type Conference or Workshop Item (Paper)
Abstract Algorithmic forecasts outperform human forecasts in many tasks. State-of-the-art machine learning (ML) algorithms have even widened that gap. Since sales forecasting plays a key role in business profitability, ML based sales forecasting can have significant advantages. However, individuals are resistant to use algorithmic forecasts. To overcome this algorithm aversion, explainable AI (XAI), where an explanation interface (XI) provides model predictions and explanations to the user, can help. However, current XAI techniques are incomprehensible for laymen. Despite the economic relevance of sales forecasting, there is no significant research effort towards aiding non-expert users make better decisions using ML forecasting systems by designing appropriate XI. We contribute to this research gap by designing a model-agnostic XI for laymen. We propose a design theory for XIs, instantiate our theory and report initial formative evaluation results. A real-world evaluation context is used: A medium-sized Swiss bakery chain provides past sales data and human forecasts.
Authors Fahse, Tobias; Blohm, Ivo; Hruby, Richard & van Giffen, Benjamin
Language English
Keywords Forecasting, Explainable AI, XAI, Design Science
Subjects business studies
information management
HSG Classification contribution to scientific community
HSG Profile Area SoM - Business Innovation
Date 18 June 2022
Publisher Association for Information Systems
Event Title European Conference on Information Systems 2022
Event Location Timisoara; Romania
Event Dates 18-24.06.2022
Contact Email Address tobias.fahse@unisg.ch
Depositing User Prof. Dr. Ivo Blohm
Date Deposited 07 Jul 2022 14:08
Last Modified 25 Sep 2022 07:12
URI: https://www.alexandria.unisg.ch/publications/266643

Download

[img] Text
Explanation Interfaces for Sales Forecasting_final.pdf

Download (1MB)

Citation

Fahse, Tobias; Blohm, Ivo; Hruby, Richard & van Giffen, Benjamin: Explanation Interfaces for Sales Forecasting. 2022. - European Conference on Information Systems 2022. - Timisoara; Romania.

Statistics

https://www.alexandria.unisg.ch/id/eprint/266643
Edit item Edit item
Feedback?