Overview

EIQAI, an expansion module of the TCalc series, is a system that fundamentally transforms EIQ analysis (E = Entry/Destinations, I = Items, Q = Quantity)—the traditional methodology used in distribution centers.
Developed as a Windows Forms application, this software embeds advanced trained models from Microsoft.ML (ML.NET) to execute high-precision AI predictions for future volume fluctuations, extending beyond static historical data aggregation.

Purpose

The primary objective of this system is to perform comprehensive logistics simulations from minimal input data and generate detailed EIQ matrices.
By specifying just two input parameters—"number of delivery destinations" and "number of items"—the system automatically computes AI predictions for 162 variable parameters required for EIQ analysis.
Specifically, it instantly calculates everything from basic indicators like loose quantities and line counts to destination ranks (E1–E5), item ranks (I1–I5), and 25 rank-based metrics covering line counts, loose items, cases, pallets, volume, and weight.
To make these extensive forecast results intuitive, the system visualizes them on-screen as EIQ matrix tables and traditional calculation views, powerfully supporting distribution center engineering and capacity planning.

Operating Environment & Tech Stack

This system adopts the following technical stack to deliver high performance and secure local execution:

Key Features

• AI Prediction Execution: Entering destination count and item count triggers 162 pre-trained models (Microsoft.ML) operating in the background to simultaneously compute AI predictions for future metric values.
• Automated EIQ Matrix Generation: Automatically builds a 5×5 EIQ matrix combining GPLE ranks (E1–E5) and GPLI ranks (I1–I5) on screen. Furthermore, users can switch display units between order frequency, loose items, cases, pallets, volume, and weight with a single click.
• Traditional Calculation E Table Output: Generates cross-sectional summary tables across destination ranks (GPLE_A1 to D) covering shipment frequency, loose item quantity, case equivalents, pallet equivalents, volume equivalents, and weight equivalents for users accustomed to traditional methods.
• Traditional Calculation I Table Output: Similarly aggregates metrics along item ranks (GPLI_A1 to D) in column directions, providing multi-angle perspectives.
• Forecast Result Data Export: Exports on-screen forecast results to detailed CSV files (PredictionResults_yyyyMMdd_HHmmss.csv) for secondary analysis.
• Comparison with Actual Calculations & Discrepancy Visualization: Verifies model accuracy by evaluating differences between historical actual calculations and current AI prediction results as cell-level, row-level, column-level, or total percentage (%) error metrics.
• Reloading Trained Models: Re-scans .zip files saved in the model folder, enabling seamless swapping to updated models without restarting the application.
• Automated Access DB Optimization: Checks DB file sizes on startup and automatically executes compact/repair operations if exceeding threshold (100MB) to maintain peak system performance.

Development Information

Overall System Flow

Below is the overall operational flow in TCalc EIQAI from application launch to EIQ matrix rendering and error verification:

Detailed Step Descriptions

【1】 Form Load (TeraCalculationEIQAI_Load)

The initialization process executed when launching the application:

• Calls FileCheck() to scan storage drives for target data.
• Initializes MLContext(seed:=0), setting up the foundation for Microsoft.ML AI predictions.
• Strictly verifies the existence of the model storage folder.
• Executes InitOutputFieldNameList() to define target fields:
  • Loose Qty, Line Count, E1–E5, I1–I5
  • Line Count 01–25, Loose Qty 01–25, Case 01–25, PL 01–25, Volume 01–25, Weight 01–25
  • Total: 162 fields
• Expands all .zip files within the model folder into memory as ITransformer instances via LoadTrainedModels().
• Calls InitAIPredictionEIQMatrixDgv() to establish the 7-column × 5-row header structure for the EIQ matrix.

【2】 Execute Prediction (ExecutePredictionBtn_Click → ExecutePrediction())

The core process running actual AI predictions triggered by user input:

• Validates text inputs in Destination Count Tb and Item Count Tb.
• Instantiates the InputData source object (Destinations, Items).
• Executes Predict methods simultaneously across all 162 loaded models (Microsoft.ML) via CreatePredictionEngine.
• Stores output forecast values into a Dictionary(Of String, Single) data structure.
• Calls DisplayPredictionResults() to convert data into DataTables bound to the DataGridView (Dgv) for UI rendering.

【3】 Build EIQ Matrix Table (AIPredictionEIQMatrixTable())

Reconstructs calculated prediction metrics into a visual EIQ matrix:

• Extracts data from Prediction Dgv and caches into a dictionary object (pred) with rounded integer values.
• Determines variable prefixes based on the active unit radio button.
• Populates extracted I1–I5 and E1–E5 values along with totals into display textboxes.
• Populates forecast values accurately into the 5×5 = 25 matrix cells of the AI Prediction EIQ Matrix Dgv.
• Calculates row sums, column sums, and grand totals for display along table borders.

【4】 Traditional Calculation E (AIPredictionTraditionalCalcE())

Re-aggregates AI prediction data across destination ranks (E) from traditional perspectives:

• Formats 9 data columns for AIPredictionTraditionalCalcDestinationDgv.
• Allocates 1 row per GPLE rank (E1–E5), summing horizontal row metrics.
• Appends a total summary row along the table footer.

【5】 Traditional Calculation I (AIPredictionTraditionalCalcI())

Re-aggregates AI prediction data across item ranks (I) from traditional perspectives:

• Configures 9 data columns for AIPredictionTraditionalCalcItemDgv.
• Allocates 1 row per GPLI rank (I1–I5), summing vertical column metrics.
• Appends a total summary row along the table footer.

【6】 Switch Display Unit (Filter_DisplayUnit_CheckedChanged)

UI handler dynamically updating EIQ matrix content:

• Executes only when internal state matrixActive = True is satisfied.
• Toggling radio buttons immediately re-runs AIPredictionEIQMatrixTable() to refresh display data.
• Users select freely among 6 units: Order Frequency, Loose Qty, Case, PL, Volume, and Weight.

【7】 Export Results (ExportResultsBtn_Click)

Export handler for external data integration:

• Exports complete prediction Dgv dataset into CSV format.
• Auto-generates filenames with timestamps: PredictionResults_yyyyMMdd_HHmmss.csv.
• Prepends baseline input parameters (Destinations, Items) as header comments in the file.

【8】 Reload Trained Models (LoadTrainedModelsBtn_Click)

• Re-scans .zip files in the model folder to dynamically reload updated AI models into memory.

【9】 Compare Predictions with Actuals (StartComparisonBtn_Click)

• Runs StartComparison() followed by DisplayErrors() to evaluate variance between forecasts and ground-truth values.

Data Structures

Data structure reference managed within TCalc EIQAI:

Primary UI Controls List

Reference list of primary UI controls deployed on screen:

Inputs

Buttons

Display Unit Radio Buttons

DataGridViews

Extracted Textboxes

Helper Functions

Internal utility functions: