A comprehensive approach to Queue Waiting Time Prediction using Tree-Based Ensembles with Data Balancing and Explainable AI
Abstract Queuing up for a service is sometimes an inevitable experience. The inefficiencies brought on by extended waiting times can be considerably decreased by precise waiting time prediction. Accurate prediction can substantially improve consumer satisfaction by reducing uncertainty. It is possib...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Springer
2025-07-01
|
| Series: | Discover Analytics |
| Subjects: | |
| Online Access: | https://doi.org/10.1007/s44257-025-00037-2 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Abstract Queuing up for a service is sometimes an inevitable experience. The inefficiencies brought on by extended waiting times can be considerably decreased by precise waiting time prediction. Accurate prediction can substantially improve consumer satisfaction by reducing uncertainty. It is possible to introduce a robust approach to the prediction of waiting times based on previous queuing data and artificial intelligence (AI) algorithms. This paper contributes to the field by offering a robust approach to waiting time prediction and suggests potential directions for further research. The investigation leverages ensemble tree-based methods along with one statistical model, supplemented by various data pre-processing techniques for regression analysis to forecast precise waiting times. The following regression models have been used to assess the performance: Random Forest (RF), Extra Trees (ET), Gradient Boosting (GBR), Histogram-Based Gradient Boosting (HGBR), Voting (VR) and Ridge Regression. Among these, the ET Regressor demonstrates superior performance. Principal Component Analysis (PCA), t-distributed Stochastic Neighbor Embedding (t-SNE), and Autoencoders have been evaluated to compare the effectiveness of different dimensionality reduction techniques. Furthermore, the challenge of data imbalance in classification tasks has also been addressed here using the Synthetic Minority Oversampling Technique (SMOTE). This process impressively enhances classification accuracy, especially for minority classes. Transparency and trustworthiness in the predictive system have been ensured through the use of Explainable Artificial Intelligence (XAI) techniques, which help interpret the decision-making processes of the models. |
|---|---|
| ISSN: | 2731-8117 |