Printing Cost Prediction Using Trained Models for Reprint Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing printing cost estimation methods in printing factories are often inaccurate, relying on rough estimates and craftsman's sense, leading to potential profit loss or order loss due to unclear cost calculations, especially for reprinted matters.
Innovation Solution
An information processing system that utilizes a trained model to predict printing costs based on input order reception information, including print job details, consumable usage, operation situation, worker information, outsourcing, and atmospheric conditions, using a neural network model for accurate cost estimation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If cost estimation is performed using rough estimates based on craftsman's sense, then the estimation process is simple and quick, but the accuracy and reproducibility of cost information are not guaranteed
Solution Approach 1:
The patent replaces the mechanical system of human craftsman estimation with an information processing system that uses trained models (neural networks or machine learning models) to automatically estimate costs. The processor inputs order reception information into the trained model, which outputs accurate cost information without relying on human sense, thereby improving accuracy while maintaining operational simplicity.
2Reliability
If accurate cost information is obtained through detailed calculation including reprinted matter, then the profit margin is improved, but the calculation process becomes complex and time-consuming
Solution Approach 1:
The patent applies preliminary action by pre-training the model using historical order reception information and corresponding cost information before actual cost estimation is needed. The trained model is stored in advance and can quickly provide accurate cost estimates for new orders without requiring complex real-time calculations, thus improving both accuracy and speed.
Solution Approach 2:
The patent uses copying by training the model on historical data (copies of past orders and costs) to learn patterns and relationships. The trained model then copies these learned patterns to estimate costs for new orders, providing accurate results without repeating the complex manual calculation processes used to generate the training data.
3Reliability
If cost estimation presents high cost information with too much margin, then the profit is protected, but the order may be lost to customers
Solution Approach 1:
The patent applies feedback by using actual cost outcomes from executed orders to continuously refine and retrain the estimation model. The model learns from the feedback of whether estimates were accurate and adjusts its predictions accordingly, providing cost information that is accurate enough to protect profits but not excessively high to lose orders.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An information processing system includes a processor configured to, in a case where order reception information related to printing, that is, the order reception information including print job information and printed matter type information is input, input new order reception information related to new printing to a trained model which has been trained in advance in order to output cost information indicating costs regarding the printing, thereby outputting new cost information regarding printing corresponding to the new order reception information.