Workflow Delay Prediction With Drift-Triggered ML Retraining
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning models for predicting delays in workflows struggle to adapt to dynamic changes in data patterns, leading to reduced accuracy and reliability due to data drift, particularly in construction and engineering projects.
Innovation Solution
A system that continuously monitors data growth and employs a combination of statistical approaches to detect data drift, triggering retraining of the ML model when necessary, ensuring the model remains adapted to changing data distributions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the ML model is continuously retrained to adapt to changing data patterns, then prediction accuracy is improved, but system complexity and computational resources increase
Solution Approach 1:
The patent implements dynamic model retraining by continuously monitoring data drift metrics and automatically triggering retraining only when drift exceeds predefined thresholds. This dynamic approach allows the system to adapt to changing data patterns while avoiding unnecessary retraining operations, thereby maintaining prediction accuracy without proportionally increasing system complexity.
Solution Approach 2:
The system incorporates feedback mechanisms through data drift detection that monitors changes in data distributions and feeds this information back to the model training process. When drift is detected, the system automatically initiates retraining with new data, creating a closed-loop feedback system that maintains accuracy while controlling complexity through condition-based activation.
2Reliability
If the ML model is retrained frequently to maintain accuracy, then prediction reliability is improved, but processing time and computational resources are consumed
Solution Approach 1:
The patent implements periodic monitoring of data drift metrics at scheduled intervals, with model retraining triggered only when drift thresholds are exceeded. This periodic approach ensures the model remains reliable by detecting changes in data patterns, while avoiding continuous retraining that would consume excessive processing time and computational resources.
Solution Approach 2:
The system changes the operational parameters of the ML model dynamically by adjusting training frequency based on detected data drift levels. When drift parameters indicate significant changes in data distributions, the system increases training frequency; otherwise, it maintains lower training frequency, thereby balancing reliability with processing time consumption.
3Adaptability or versatility
If statistical approaches are used to detect data drift, then model adaptability is improved, but computational overhead increases
Solution Approach 1:
The patent applies statistical approaches partially by using them only for drift detection rather than for the entire model training process. This selective application provides sufficient adaptability to detect when retraining is needed while minimizing computational overhead by avoiding unnecessary statistical computations during periods of stable data patterns.
Data Source
AI summary
According to an aspect, a system collects historical data indicating details of multiple closed workflows and trains an ML model based on the multiple closed workflows, the ML model thereafter operable to predict delays for open workflows. Upon receiving, after the training, details of an additional set of closed workflows, the system adds the received details to the historical data to form an updated historical data. The system checks whether the updated historical data has a data growth (in comparison to the historical data) exceeding a threshold. If the data growth exceeds the threshold, the system determines whether there exists a data drift in the updated historical data in comparison to the historical data. If the data drift exists, the system retrains the ML model based on the updated historical data, wherein the retrained ML model is thereafter operable to predict delays for open workflows.