Missing Values and Optimal Selection of an Imputation Method and Classification Algorithm to Improve the Accuracy of Ubiquitous Computing Applications

Table 12

Factors influencing accuracy (RMSE) of classifier algorithms.

Data characteristic		Data characteristic
(constant)	.060^**	M_imputation_dum1	.012^**
R_missing	.083^**	M_imputation_dum2	−.001^*
SE_HS	−.005^**	M_imputation_dum4	000
SE_VS	.000^**	M_imputation_dum5	000
Spread	.017^**	M_imputation_dum6	.001^**
N_attributes	−.008^**	M_imputation_dum7	−.001^*
C_imbalance	−.003^**	P_missing_dum1	−.006^**
N_cases	.002^**	P_missing_dum3	.000

Note  1: Dummy variables related to imputation methods: LISTWISE DELETION (M_imputation_dum1 = 1, others = 0), MEAN_IMPUTATION (M_imputation_dum2 = 1, others = 0), GROUP_MEAN_IMPUTATION (M_imputation_dum3 = 1, others = 0), PREDICTIVE_MEAN_IMPUTATION (M_imputation_dum4 = 1, others = 0), HOT_DECK (M_imputation_dum5 = 1, others = 0), -NN (M_imputation_dum6 = 1, others = 0), and -MEANS_CLUSTERING (M_imputation_dum7 = 1, others = 0). Missing patterns: univariate (P_missing_dum1 = 1, P_missing_dum2 = 0, P_missing_dum3 = 0), monotone (P_missing_dum1 = 0, P_missing_dum2 = 1, P_missing_dum3 = 0), and arbitrary (P_missing_dum1 = 1, P_missing_dum2 = 1, P_missing_dum3 = 1). : standard beta coefficient.
Note  2: ^* < 0.1, ^** < 0.05.