EIQ_AI Model Creation
AI Prediction Model Creation, Training, and Prediction System Based on EIQ Analysis
VB.NET
.NET Framework 4.8
Microsoft.ML
Access (.accdb)
1. System Overview
This system is a Windows Forms application that constructs AI prediction models from shipping data using the EIQ (Entry-Item-Quantity) analysis method.
By using the number of Destinations (E) and Items (I) as input, it predicts approximately 160 items—including piece quantity, number of lines, EIQ indicators, case counts, pallet counts, volume, and weight—using SDCA Regression (Microsoft.ML).
Key Features
- Shipping Data Pre-processing — Importing Excel data, unit conversion, and ABC ranking.
- Model Data Configuration — Setting E/I ratio ranges and record pitch, plus progress management.
- Model Data Creation — Generating the T600 data model table based on settings.
- Machine Learning — CSV export, training, prediction, and result CSV output.
【Technical Insight: From Static Analysis to Dynamic Prediction】
Traditional EIQ analysis was limited to "summarizing past performance (static points)". In this system, AI learns fluctuations in the number of Destinations (E) and Items (I) to evolve EIQ analysis into a "prediction model" that dynamically simulates facility capacity and workload. This system is a next-generation simulation foundation that algorithmizes 30 years of logistics consulting expertise. While traditional EIQ analysis merely aggregated past results, this system predicts dynamic capacity for new or integrated facilities, serving as a "Logistics Digital Twin".
2. File Configuration
| File | Class / Module | Role |
|---|
E0_Menu.vb | E0_MenuF | Main menu. Transitions to each process screen and checks DB status. |
E1_DataPre-processing.vb | EIQ_AI_Pre-processingF | Imports shipping data, unit conversion, EIQ ranking, and T200 table creation. |
E2_ModelConfig.vb | ModelDataConfigF | E/I ratio and pitch configuration, T300 progress table management. |
E3_ModelCreation.vb | E3_ModelCreationF | Creates the T600 data model based on T300 progress (batch processing). |
E4_MachineLearning.vb | E4_MLF | Model training, prediction execution, and result export via Microsoft.ML. |
EIQ_AI_ProValidation.vb | EIQ_AI_ProValidation | AI verification and validation (for future expansion). |
K1.vb | Module K1 | Global variables, constants, and file path management. |
AccessDbManager.vb | AccessDbManager | Access database creation and Excel import. |
AccessOps.vb | — | SQL execution helper, common table operation processing. |
TableStructureCheck.vb | — | Diagnostic utility for table structure verification. |
【Practical Significance: Data Cleansing and Feature Engineering】
Raw logistics data often contains inconsistent units or errors; thus, "Automated Unit Conversion" aligns shipping units such as pieces and cases to ensure calculation accuracy. Additionally, "5-Stage Pareto Analysis (EIQ Ranking)" is automatically executed to let the AI recognize which product groups are primary drivers of center load as "features". Built on VB.NET and Microsoft Access, it ensures the confidentiality of sensitive data by processing it in-house without external cloud dependencies.
3. Overall Process Flow
▼
Route A
Create from Data
Import Excel Files
Route B
Attach to Tera-Calc
Link existing T200 DB
▼
E1 Data Pre-processingDB Creation → Import → Unit Conversion → T200 Creation → Ranking
▼
E2 Model ConfigurationE/I ratio (50-150%) and pitch configuration → T300 Creation
▼
E3 Model CreationGenerate T600 model for T300 records (Supports pause/resume)
▼
E4 Machine LearningLoad CSV → Train SDCA Regression (approx. 160 models) → Predict → Export
▼
EIQ_AI VerificationValidation of trained models vs. real data (Future expansion)
Generating thousands of "virtual shipping scenarios" by varying E/I ratios from 50% to 150% to educate the AI.
4. Each Step Details
E1 Data Pre-processing EIQ_AI_Pre-processingF
Processes shipping data from import to the creation of the T200 table for EIQ analysis.
ShippingData.xlsx →
DB Creation →
T000_RawData →
T100_Data →
T110_Conversion →
T180_EIQRank →
T200 (Integrated)
📌 ABC Rank Criteria: 5-stage ranking (A1/A2/B/C/D) based on Pareto analysis of cumulative ratios.
E2/E3 Model Data Creation
| Parameter | Range | Description |
|---|
| E Min/Max Ratio | 50% 〜 150% | Range of Destination count variation |
| I Min/Max Ratio | 50% 〜 150% | Range of Item count variation |
| Record Pitch | 1 〜 20 (Default 5.9) | Interval for generated records |
E4 Machine Learning E4_MLF
DataModel.csv →
Feature Mapping (E, I) →
MinMax Normalization →
SDCA Training →
.zip Model Save
5. Database Structure
Database: EIQ_AI_Model.accdb (Microsoft Access)
| Table | Origin | Purpose |
|---|
T200 | E1 | EIQ processed data (The foundation) |
T300_Progress | E2 | Management of simulation ranges and creation flags |
T600_DataModel | E3 | AI training model data (165 fields) |
6. Execution Environment & Prerequisites
| Item | Requirement |
|---|
| Framework | .NET Framework 4.8 |
| Libraries | Microsoft.ML, ClosedXML, ADOX |
| DB Engine | ACE OLEDB 12.0 (Microsoft Access) |
7. Data Transformation Flow Chart
Raw Excel →
T200 (Structuring) →
T600 (Simulation Space) →
🤖 AI Prediction Model
EIQ_AI Model Creation — README_EN.html
By simulating extreme conditions—such as "What if destinations increase by 1.5x?" or "What if items are halved?"—the AI can accurately grasp the facility's "growth potential" and "operational limits".