K-nearest neighbour (KNN) algorithm in combination with a genetic algorithm were applied to a medical fraud detection problem. The genetic algorithm was used to determine the optimal weighting of the features used to classify General Practitioners' (GP) practice profiles. The weights were used in the KNN algorithm to identify the nearest neighbour practice profiles and then two rules (i.e. the majority rule and the Bayesian rule) were applied to determine the classifications of the practice profiles. The results indicate that this classification methodology achieved good generalisation in classifying GP practice profiles in a test dataset. This opens the way towards its application in the medical fraud detection at Health Insurance Commission (HIC).
[1]
John H. Holland,et al.
Adaptation in Natural and Artificial Systems: An Introductory Analysis with Applications to Biology, Control, and Artificial Intelligence
,
1992
.
[2]
Hongxing He,et al.
Application of neural networks to detection of medical fraud
,
1997
.
[3]
Belur V. Dasarathy,et al.
Nearest neighbor (NN) norms: NN pattern classification techniques
,
1991
.
[4]
John H. Holland,et al.
Adaptation in Natural and Artificial Systems: An Introductory Analysis with Applications to Biology, Control, and Artificial Intelligence
,
1992
.