Important Neighbors: A Novel Approach to Binary Classification in High Dimensional Data

Table 2

Comparison of methods in terms of probability of achieving the maximum accuracy (PAMA) and probability of achieving more than 95% of the maximum accuracy (P95).

Degree of sparsity	Method	PAMA	P95
90%	Random forest (RF)	45.8%	66.7%
	Support vector machine (SVM)	50.0%	87.5%
	nearest neighbors (KNN)	0.0%	20.8%
	important neighbors (KIN)	8.3%	41.7%
95%	Random forest (RF)	50%	79.2%
	Support vector machine (SVM)	29.2%	79.2%
	nearest neighbors (KNN)	0.0%	29.2%
	important neighbors (KIN)	20.8%	75%
98%	Random forest (RF)	20.8%	75%
	Support vector machine (SVM)	12.5%	41.7%
	nearest neighbors (KNN)	0.0%	33.3%
	important neighbors (KIN)	66.7%	100%
Total	Random forest (RF)	38.9%	73.6%
	Support vector machine (SVM)	30.6%	69.4%
	nearest neighbors (KNN)	0.0%	27.8%
	important neighbors (KIN)	32%	72.2%