Using autoencoders as a weight initialization method on deep neural networks for disease detection

Table 10 Performance comparison of the classifier, for the Basic AE, when changing the imputation strategy at the data preprocessing step

Strategy	Top Layers (AEs)	Accuracy (%)	MCC	Precision (%)	Recall (%)	F₁ score
Fixing the AE weights (Approach A)
Mean ^∗	AE: Encoding Layers	88.40 ±5.52	0.59 ±0.17	68.39 ±19.13	64.80 ±10.84	65.91 ±13.72
	AE: Complete AE	91.77 ±3.13	0.69 ±0.12	80.57 ±11.79	67.00 ±11.24	72.91 ±10.86
CV	AE: Encoding Layers	91.93 ±2.13	0.69 ±0.10	79.43 ±6.20	69.40 ±10.96	73.81 ±8.14
	AE: Complete AE	93.23 ±1.99	0.74 ±0.08	83.41 ±5.85	74.20 ±9.59	78.31 ±6.95
MFV	AE: Encoding Layers	92.50 ±2.36	0.71 ±0.10	82.60 ±7.41	70.00 ±13.40	75.16 ±9.23
	AE: Complete AE	93.27 ±1.71	0.74 ±0.07	84.97 ±4.01	72.40 ±9.74	77.91 ±6.54
Fine-Tuning the AE Weights (Approach B)
Mean ^∗	AE: Encoding Layers	99.33 ±0.52	0.98 ±0.02	97.85 ±2.32	98.20 ±1.48	98.01 ±1.55
	AE: Complete AE	99.30 ±0.37	0.98 ±0.01	99.00 ±1.06	96.80 ±2.35	97.87 ±1.15
CV	AE: Encoding Layers	99.40 ±0.49	0.98 ±0.02	98.63 ±2.04	97.80 ±2.39	98.23 ±1.48
	AE: Complete AE	99.30 ±0.53	0.98 ±0.02	99.01 ±1.39	96.80 ±3.29	97.97 ±1.38
MFV	AE: Encoding Layers	99.47 ±0.32	0.98 ±0.01	98.83 ±1.64	98.00 ±2.11	98.39 ±0.98
	AE: Complete AE	99.13 ±0.57	0.97 ±0.02	98.77 ±1.71	96.00 ±2.31	97.36 ±1.74

The experiment pipeline remains the same, under the same evaluation metrics. The Strategy column represents the imputation strategy used. The ∗ symbol represents the default strategy. The following abreviations were used: CV for Constante Value, and MFV for Most Frequent Value

ISSN: 1472-6947