A causal model for short‐term time series analysis to predict incoming Medicare workload. (20th July 2020)

• Record Type: Journal Article
• Authors: Mizan, Tasquia; Taghipour, Sharareh
• Is Part Of: Journal of forecasting. Volume 40:Number 2(2021)
• Page Start: 228
• Page End: 242
• Publication Date: 2020-07-20
• Subjects: batch training -- causal model -- ensemble of experts -- short‐term time series -- workload prediction; Forecasting -- Periodicals; Forecasting -- Mathematical models -- Periodicals; 003.2
• DOI: 10.1002/for.2717
• Languages: English
• ISSNs: 0277-6693

Abstract

Abstract: We have investigated methodologies for predicting radiologists' workload in a short time interval by adopting a machine learning technique. Predicting for shorter intervals requires lower execution time combined with higher accuracy. To deal with this issue, an ensemble model is proposed with the fixed‐batch‐training method. To excel in the execution time, a fixed‐batch‐training method is used. On the other hand, the ensemble of multiple machine learning algorithms provides higher accuracy. The experimental result shows that this predictive model can produce at least 10% higher accuracy in comparison with the other available widely used short‐term time series forecasting models. In the studied medical system, this gain in accuracy for the earlier prediction of workload can reduce the Medicare relative value unit cost by $1.1 million annually, which we have formulated and shown in this paper. The proposed batch‐trained ensemble of experts model has also provided at least a 6% improvement in execution time compared with the other studied models.