Malignant Mesothelioma Disease Diagnosis using Data Mining Techniques

Figures & data

Figure 1. The CRISP-DM tool.

Figure 2. MLP neural network (left) and SVM (right).

Table

Total population	Predicted condition positive	Predicted condition negative
Condition positive	True positive (TP)	False negative (FN)
Condition negative	False positive (FP)	True negative (TN)

Table 1. Comparison between SVM and MLPE methods for MM disease diagnosis in terms of average performance evolutions by 10-fold crossvalidation in 5-run test methods.

Performance evolution
Methods	ACC (%)	AUC (0–1)	TPR (%)	TNR (%)	F1(%)
SVM	99.87	0.9999	99.8245	100	99.9560
MLPE	99.56	0.9998	99.5614	98.5416	99.6485

SVM: Support vector machine, MLPE: Multilayer perceptron ensembles, ACC: Classification accuracy rate, AUC: Area under curve, TPR: True positive rate, TNR: True negative rate, F1: F1 score.

Table 2. Average of classification accuracies for MM disease dataset by 10-fold crossvalidation in 5 runs.

Results (%) per run (average of 10-fold crossvalidation outputs in each run)
Methods	1^s	2^nd	3^rd	4^th	5^th	Average
SVM	100	100	99.69	100	99.69	99.87
MLPE	99.38	99.35	99.69	99.69	99.69	99.56

SVM: Support vector machine, MLPE: Multilayer perceptron ensembles

Table 3. Comparison of different methods used to measure the performance evolution for MM disease diagnosis in terms of average classification accuracy.

Methods	Number of fold crossvalidation	Number of runs	ACC (%)
SVM*	10	5	99.87
MLPE*	10	5	99.56
AIS	10	—-	97.70
NN	10	—-	91.30
PNN	3	—-	96.30
MLNN	3	—-	94.41
LVQ	3	—-	91.14

‘*’: Proposed methods in this study, AIS: Artificial immune systems, NN: Neural network, PNN: Probabilistic neural network, MLNN: Multilayer neural network, LVQ: Learning vector quantization.

Figure 3. LIFT plot for MM disease diagnosis using SVM and MLPE neural network.

Figure 4. Input importance bar charts for MM disease diagnosis using 34 input variables.

Figure 5. Variable effect curve for the input variable ALP.