SCO Terminal Failure Prediction With Adaptive Model Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The increasing adoption of Self-Checkout (SCO) lanes in retail stores due to the COVID-19 pandemic has led to a rise in equipment failures, causing customer dissatisfaction, increased costs for retailers, and potential loss of revenue, as well as prolonged downtime due to the lack of skilled staff to resolve issues promptly.
Innovation Solution
A predictive maintenance system using multiple machine learning models trained on terminal data to provide real-time failure predictions, selecting the highest-scoring model for deployment in a production environment, allowing for proactive maintenance and adaptation to changing patterns in terminal data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple machine learning models are trained and tested to select the highest-scoring model for deployment, then prediction accuracy is improved, but system complexity and training time increase
Solution Approach 1:
The patent applies preliminary action by training and evaluating multiple machine learning models in advance during an experimentation phase before deployment. Different model types (e.g., XGBoost, LightGBM, CatBoost, neural networks) are pre-trained on historical data, and their performance is evaluated using metrics like AUC-ROC and F1-score. The best-performing model is selected ahead of time for production deployment, ensuring high prediction accuracy without requiring complex real-time model comparison during actual failure prediction operations.
2Reliability
If predictive maintenance is implemented to reduce SCO downtime, then customer satisfaction is improved, but implementation costs and system complexity increase
Solution Approach 1:
The patent applies self-service by implementing an automated predictive maintenance system that continuously collects terminal data, processes it through selected machine learning models, and generates failure predictions without requiring constant human intervention. The system autonomously monitors terminal health metrics, identifies patterns indicating potential failures, and alerts maintenance personnel in advance, allowing the system to serve itself in detecting and reporting issues while reducing overall operational complexity.
3Productivity
If real-time failure predictions are provided using deployed models, then operational disruptions are reduced, but computational resources and processing time increase
Solution Approach 1:
The patent applies preliminary action by pre-training multiple machine learning models on historical terminal data during an experimentation phase before deployment. The selected model is pre-optimized and ready for production use, allowing real-time failure predictions to be generated efficiently without requiring extensive computational resources during operational execution. This advance preparation reduces the computational burden during actual real-time prediction operations.
Data Source
AI summary
Multiple types of machine learning models (MLMs) are trained on training data to provide predictions of terminal failures. The MLMs are tested on testing data and prediction scores are calculated for the MLMs based on the testing. An MLM having a highest prediction score is selected and deployed to a production environment for a current period of time to provide predictions of terminal failures. Current terminal data is provided as input to the deployed MLM, and the MLM's predictions are provided to an interface, system, or service of the production environment. The current terminal data is also provided to the non-selected MLMs, which generate candidate predictions during the current period of time. At the end of the current period of time, prediction scores of the selected and non-selected MLMs are calculated. A next MLM is selected for a next period of time for deployment based on the prediction scores.