| < Day Day Up > |
|
Discriminant analysis yielded 50.4% of original grouped cases correctly classified. No significant relationship was found (.149) between attitude, morale, internal controls, increases in expenditures and whether or not fraud was actually committed. Cronbach's alpha of reliability was .6626 and offered somewhat reliable results in this exploratory research.
Neural networks did a much better job of predicting fraud (75.9%) good parts than discriminant analysis (50.4%). Neural networks were able to find patterns in the training set and then correctly identify more than three fourths of similar patterns in the testing set. Therefore, it can be concluded that neural networks outperform discriminant analysis by 25.5% in this data set.
| < Day Day Up > |
|