Parallel Computing Framework for ETA Determination
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for determining the estimated time of arrival (ETA) in on-demand transportation services are slow and inaccurate, failing to provide timely and reliable estimates.
Innovation Solution
A system utilizing a parallel computing framework with worker nodes trained using decision tree models, such as Extreme Gradient Boosting (XGboost), to generate a global ETA model by allocating feature data across multiple worker nodes, enabling faster and more accurate ETA calculations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional single-node computing methods are used to determine ETA, then the system complexity is low, but the computation speed is slow and accuracy is insufficient
Solution Approach 1:
The patent divides the computing system into multiple worker nodes (first worker node, second worker node, etc.), where each node is responsible for training specific sub-ETA models. This segmentation allows parallel processing of different feature data subsets, significantly improving ETA determination speed while distributing system complexity across multiple independent nodes rather than concentrating it in a single complex system.
Solution Approach 2:
The patent transitions from a single-node computing architecture to a multi-node parallel computing architecture, adding the dimension of spatial distribution. By allocating different sub-ETA models to different worker nodes and combining their results, the system achieves faster computation without proportionally increasing overall complexity through modular design.
2Measurement precision
If traditional computing methods are used, then the system structure is simple, but the ETA accuracy is insufficient
Solution Approach 1:
The patent segments the ETA determination task into multiple sub-ETA models (first sub-ETA model, second sub-ETA model, etc.), with each model trained on specific feature data by dedicated worker nodes. This segmentation allows each sub-model to specialize in particular aspects of ETA prediction, improving overall accuracy while distributing model training complexity across multiple nodes rather than requiring one extremely complex monolithic model.
Solution Approach 2:
The patent combines the predictions from multiple sub-ETA models to generate the final global ETA. By merging the results from different worker nodes that each contribute specialized sub-models, the system achieves higher accuracy through ensemble prediction while managing complexity through modular architecture where each node handles a manageable subset of the overall problem.
Data Source
AI summary
The present disclosure relates to methods and systems for determining an estimated time of arrival (ETA). The methods may include obtaining feature data related to an on-demand service order; obtain a parallel computing framework; determining a global ETA model based on the feature data and the parallel computing framework; and determining an ETA for a target route based on the global ETA model.