Understanding Incorrect% in a PNN Testing Report

Applies to: NeuralTools PNN Testing Reports

 

What is Incorrect%?

• Incorrect% is the sum of the probabilities for all incorrect categories in a PNN testing report.

• Formula:

[
\text{Incorrect%} = 100% - \text{Probability of Correct Category}
]

How to View Probabilities of All Categories

1. Go to Utilities » Application Settings.

2. Under Reports, select Columns in Detailed Reports.

3. Enable Probabilities of All Categories (for PNN) for Testing.

Note: Ensure your dataset has at least three dependent variable categories to see meaningful results (e.g., Auto Loans 1a example).

 

Example from a Detailed Testing Report

• Row 5:

  • Correct category = “late payments” → 43.12%

  • NeuralTools predicted:

    • “timely payments” → 54.69%

    • “default” → 2.19%

• Calculation:

[
\text{Incorrect%} = 100 - 43.12 = 56.88%
]

• This matches the sum of the incorrect probabilities: 54.69 + 2.19 = 56.88%.

Good vs. Bad Results

• Good results: Prediction% = correct%, so Incorrect% = 100 – Prediction%

• Bad results: Prediction% shows the probability for the predicted (wrong) category, and Incorrect% sums all incorrect category probabilities

 

Tip: Always check the Probabilities of All Categories column to verify how NeuralTools distributes predictions across all possible outcomes.

In row 5, the correct category was “late payments” (43.12%), but NeuralTools predicted “timely payments” (54.69%), with “default” at 2.19%. Thus, Incorrect% = 100 – 43.12 = 56.88%, matching the sum of the incorrect probabilities (54.69 + 2.19).

For both Good and Bad results, Incorrect% = 100 – correct%. NeuralTools shows Prediction%, not correct%—for Good results, they are the same; for Bad results, Prediction% is for the wrong category, and Incorrect% sums all incorrect probabilities.