AI Auto Modeling
If you have an Excel file in hand but are not sure how to configure it——which row is the title, which columns are features, and which are targets——AI Auto Modeling is made for you.
It hands the entire chain that used to be done manually——"read the structure → come up with a name → configure variables → create the model → run the fitting → lay out the charts"——to the AI; all you need to do is drag in a file + choose a modeling type.
🚪 Entry
Top toolbar → Add Instance → switch to the "AI Build" tab
💡 On the right side of the dialog there is a permanent instructions and sample download panel, which explains the differences between the three modeling types and provides sample data you can download directly (PCA sample, PLS sample, PLS-DA sample).
⚠️ Prerequisites
| Prerequisite | Description |
|---|---|
| A licensed version other than the free version | The free version shows "The current license version does not support AI Build mode" and requires a license upgrade |
| The large language model has been configured in Settings | A valid API Key is required, otherwise structure identification cannot be performed |
| The file is in Excel format | Supports .xlsx / .xls |
🧩 Three Modeling Types
AI Build offers the three most commonly used modeling types (to use OPLS / OPLS-DA, go through the manual modeling flow):
| Option | Applicable Scenario | Is a Y Column Needed |
|---|---|---|
| Analysis (PCA) | You only want to look at the data structure and find abnormal batches, with no target variable | Not needed |
| Regression (PLS) | Y is a continuous value (such as purity, yield, strength) | Needed |
| Classification (PLS-DA) | Y is a class label (such as pass/fail, grade A/B/C) | Needed |
💡 Not sure which one to choose? First check whether your Excel has the column you want to predict: no → PCA; yes and it is numeric → PLS; yes and it is text/a category → PLS-DA.
🔄 Four-Step Build Flow
After clicking "Start AI Build", the interface shows a circular progress bar and real-time status text:
Step 1: Parse the file locally and extract data features
- Read the Excel with the browser engine locally, extracting the first 10 rows as a data preview
- This step is offline; the data never leaves your computer
- Empty rows at the end of the file are skipped automatically
Step 2: The AI identifies the data structure and generates project information and model configuration
The file name + data preview are sent to the large language model you configured, and the AI determines:
| What the AI Identifies | Description |
|---|---|
| Instance name | Generated automatically according to the rule [modeling type]-[core word of the file name], such as PCA-BloodSample |
| Instance remark | A brief description of the data and the goal |
| Title row | The row containing the English names of the points/variables (usually row 1) |
| Name row | The row containing the Chinese descriptions of the variables (usually row 2, or empty if there is none) |
| Observation column | The sample identifier column, such as time, id, a batch number (usually the first column) |
| X columns (features) | Numeric columns other than the observation column and the Y column |
| Y column (target) | Empty for a PCA model; for PLS / PLS-DA the target column is inferred by the AI |
⚠️ If the AI cannot find the title row or the observation column, it aborts the build and gives an error message; in that case it is recommended to switch to manual configuration.
Step 3: Create the instance and save the model configuration
Automatically create the instance and the model, and write the row/column configuration identified in step 2 into the model.
Step 4: Train the model and generate report charts
- Automatically perform cross-validation to determine the best number of components (
Training: component N...is displayed during training) - After training is complete, automatically lay the common charts onto the canvas (
Generating chart: <chart type>...is displayed) - When everything is finished, it shows "AI build complete!"
📋 Complete Status Message Reference
During the build, the following texts appear in sequence below the progress bar, making it easy to tell which step you are stuck on:
| Progress | Status Text |
|---|---|
| 0% | Step 1: Parsing the file locally and extracting data features... |
| 20% | Step 2: The AI is identifying the data structure and generating project information and model configuration... |
| 40% | Step 3: Creating the instance and saving the model configuration... |
| 66% | Step 4: Training the model and generating report charts... |
| 100% | AI build complete! |
🎯 When to Use It / When Not To
✅ Suitable for using AI Build:
- You are using the platform for the first time and are unfamiliar with the concepts of "title row / name row / observation column"
- You have received a set of data from someone else and are unsure of the semantics of the columns
- You want to quickly lay out a version of the charts to see what the data looks like
- You need to process multiple data files with a similar structure in batch
⚠️ Manual configuration is recommended instead:
- The data structure is special (multiple worksheets, headers not in the first two rows, merged cells)
- You need to use OPLS / OPLS-DA modeling
- You need precise control over the selection of X/Y columns (the AI's inference needs human review)
- The data is confidential and no content may be sent to an external large language model
💡 Important: What the AI identifies is the structure, not business causality. After building the model, be sure to open the data configuration and review whether the choice of X/Y columns matches business common sense before moving on to formal analysis.
🔗 Related Reading
- Step 1: Data Preparation —— Full instructions for manually configuring the data structure
- Step 2: Modeling Analysis —— Manually fitting a model and tuning parameters
- AI Capabilities Overview —— Learn about all the AI features available after modeling