Weka报错‘class attribute is not nominal!’：乳腺肿块数据集处理咨询

Hey there, let's tackle that Weka error you're facing. The "class attribute is not nominal!" message means Weka expects your target class column to be a nominal (categorical) type, but it’s currently being recognized as a numeric/integer type instead. Here’s how to fix it step by step:

Step 1: Check the Class Attribute’s Type #

First, open your dataset in Weka Explorer and navigate to the Preprocess tab:

• Locate the column you’ve set (or want to set) as your class attribute—based on your dataset details, this is likely the BI-RADS assessment column.
• Look at the "Type" field for that column. If it shows Numeric or Integer instead of Nominal, that’s exactly the problem.

Step 2: Convert the Attribute to Nominal #

You have two simple ways to adjust this:

Option 1: Modify Directly in Weka #

1. Click the Edit button next to the problematic column.
2. In the attribute editor window, switch the "Type" dropdown from Numeric to Nominal.
3. For ordinal attributes like BI-RADS (1-5), make sure the "Values" list is ordered correctly (1,2,3,4,5)—Weka will respect this sequence for algorithms that account for ordinality.
4. Click OK to save the change, then use the Save button in the Preprocess tab to store the updated dataset (you can keep it as CSV or switch to ARFF for more control over attribute definitions).

Option 2: Edit the Dataset File Directly (ARFF Format) #

If you prefer working with the raw file:

1. Import your CSV into Weka, then save it as an ARFF file using the Save button in the Preprocess tab.
2. Open the ARFF file in a text editor.
3. Find the line defining your class attribute—for BI-RADS, it might look like:

   @attribute BI-RADS numeric
   

   Replace it with:

   @attribute BI-RADS {1,2,3,4,5}
   

4. Save the ARFF file and reimport it into Weka. The attribute will now be recognized as nominal.

Step 3: Reassign the Class Attribute #

After converting the type, go back to the Preprocess tab:

• Select your target class column.
• Click the Class button in the top-right corner to confirm it’s set as the class attribute.

That should clear up the error! A quick reminder: Weka treats ordinal categorical values (like your BI-RADS 1-5) as nominal by default, but keeping the value order correct ensures any algorithms that respect ordinality will work as intended.

内容的提问来源于stack exchange，提问作者Daniel