Table 7.

A ranking list of 20 best-performing sets of hyperparameters (regarding the largest F1 score value obtained) used to perform the 5 × 2CV procedure using the k-nearest neighbor classifier. Sets of hyperparameters are shown on the left side of the table, and statistical measures describing the level and variation of performance measured by the F1 score are shown in the right part of the table

Pre-processing type and hyperparameters				Achieved F1 score statistics
d	Normalization type	k	$p$	Min	Mean	Max	Std. dev	Conf. int. (95%)
{4,8}	None	41	1.5	0.557	0.617	0.676	0.059	(0.500; 0.733)
{4,8}	None	41	1.75	0.526	0.590	0.654	0.064	(0.464; 0.716)
{4,8}	None	28	1.25	0.569	0.609	0.648	0.040	(0.531; 0.687)
{4,8}	None	41	1.25	0.560	0.602	0.645	0.043	(0.519; 0.686)
{4,8}	None	41	2	0.509	0.575	0.641	0.066	(0.446; 0.704)
{4,8}	Standardization	28	1	0.598	0.617	0.637	0.020	(0.579; 0.656)
{4,8}	None	41	1	0.594	0.615	0.636	0.021	(0.574; 0.656)
{2,6}	None	16	2	0.562	0.598	0.635	0.036	(0.528; 0.669)
{4,8}	Standardization	28	1.25	0.605	0.617	0.630	0.012	(0.593; 0.642)
{4,8}	Standardization	41	1.75	0.570	0.600	0.629	0.029	(0.542; 0.657)
{3,8}	None	16	1.25	0.602	0.615	0.628	0.013	(0.589; 0.640)
{4,8}	Standardization	41	1.5	0.556	0.589	0.622	0.033	(0.524; 0.654)
{3,8}	Standardization	16	1.75	0.539	0.580	0.622	0.041	(0.499; 0.661)
{4,8}	None	28	1.5	0.583	0.602	0.621	0.019	(0.565; 0.640)
{3,8}	None	28	1.25	0.578	0.599	0.619	0.020	(0.559; 0.639)
{3,8}	None	16	1	0.597	0.608	0.619	0.011	(0.587; 0.630)
{4,8}	Standardization	28	1.5	0.572	0.595	0.617	0.022	(0.551; 0.638)
{4,8}	None	28	1.75	0.568	0.592	0.616	0.024	(0.545; 0.639)
{3,8}	Standardization	16	1	0.559	0.587	0.615	0.028	(0.532; 0.642)
{2,7}	None	16	1.5	0.560	0.587	0.615	0.028	(0.533; 0.641)