Chapter 5, Problem 1P

Explanation of Solution

Given: A routine is applied with the help of data mining to the transaction dataset. The transaction dataset is categorized into fraudulent and non-fraudulent that has 88 records (30 correctly) and 952 records (920 correctly), respectively.

To find: The classification matrix and determine the error rate.

Solution:

In the given context, classification matrix provides the most accurate measure from the dataset.

The number of C0,1 classified cases correctly is n(0,0) = 30.

The number of C0,1 classified cases incorrectly as C1,0 is n(0,1) = 58.

The number of C1,0 classified cases incorrectly as C0,1 is n(1,0) = 32.

The number of C1,0 classified cases correctly is n(1,1) = 920.

The following is the classification matrix:

By analyzing the classification matrix, the values of both fraudulent and non-fraudulent are as follows:

The actual rate of fraudulent is n(0,1) = 58.

The actual rate of non-fraudulent is n(1,0) = 32.

Total number of records present in the dataset is n = 30 + 58 + 32 + 920 = 1040.

Evaluate the rate of error present in the dataset by using the formula of error rate.

Hence, the error rate is 8.65%.