Data Table
📋 What is this?
It is the most "plain" of all charts, but also the most irreplaceable one: it presents the dataset the model is currently using in table form, exactly as it is.
When a chart tells you "this point is abnormal", the table tells you "exactly which row this point is, and what each variable's value is".
🧐 How to read?
Row labels
Each row has a small label on the right, indicating which dataset the row belongs to:
| Label | Meaning |
|---|---|
| TR | Training set (Train) |
| TE | Test set (Test) |
💡 If a sample's prediction performance is very poor, first check whether it is TR or TE——a failure on the test set is what truly deserves attention.
Value highlighting
- Rows selected by the lasso in the chart are marked with a red / blue background
- When "Show Predicted Values" is turned on, the prediction columns are distinguished with a special style
🛠️ How to use?
The top toolbar provides five groups of controls:
1. Dataset filtering
| Option | Displayed content |
|---|---|
| All Data | Training set + test set |
| Training Set Only | Only TR rows |
| Test Set Only | Only TE rows |
2. Selected row filtering
Used together with the lasso or Shift / Ctrl + click:
| Option | Displayed content |
|---|---|
| Red Selected Rows | The red group you circled in the chart |
| Blue Selected Rows | The blue group you circled in the chart |
💡 This is the best way to verify whether your circling was accurate: after circling, switch to "Red Selected Rows" and check row by row for anything you circled by mistake.
3. Header display toggle
| Toggle | Effect |
|---|---|
| Show Alias | The header shows the variable's Chinese name |
| Show Full Name | The header shows the variable's English point name |
4. Predicted value toggle
- Show Predicted Values —— display the model prediction next to the original Y column for direct comparison
- Hide Predicted Values —— only look at the raw data
5. Search sample ID
A search box is provided at the top right; enter a sample ID to quickly locate a row.
💡 Practical use: when the AI cleaning report calls out "sample 3394 is abnormal", just enter
3394in the search box to jump to that row and inspect all its variable values.
🎯 Typical uses
- Verify abnormal samples: get the sample ID from the AI report → search and locate → check column by column
- Verify selection results: switch to "Red Selected Rows" to confirm the grouping is correct
- Check data quality: scan the raw values with your eyes to spot unit errors and typos
- Compare prediction with actual: turn on "Show Predicted Values" and find the rows with the largest deviation
⚠️ The amount of data the table displays is limited by the current model's dataset range; if you need the complete raw data, go to Instance List → Download to export a CSV.