From: Comparing different supervised machine learning algorithms for disease prediction
Criteria | # articles meet this criterion (%) | Name and frequency of the algorithm that showed ‘superior’ accuracy | |
---|---|---|---|
Most times | Second most times | ||
Disease names that were frequently modelled | |||
 Heart disease | 23 (48%) | NB, SVM (4 times, each) | ANN, DT, KNN, LR (3 times, each) |
 Diabetes | 7 (15%) | SVM (4 times) | RF (2 times) |
 Breast cancer | 5 (10%) | ANN (2 times) | DT, RF, SVM (1 time, each) |
 Parkinson’s disease | 3 (6%) | SVM (2 times) | KNN (1 time) |
Type of the data that were used | |||
 Clinical and Demographical | 15 (30%) | DT (6) | ANN, KNN, NB, RF (2 times, each) |
 Other data types | 33 (66%) | SVM, RF (12 times, each) | RF (7) |
Validation method followed | |||
 10-fold validation | 21 (42%) | SVM (5 times) | DT, RF (4 times, each) |
 5-fold validation | 6 (12%) | SVM (3 times) | RD (2 times) |
 Other method | 7 (14%) | LR, NB, SVM (2 times, each) | DT (1 time) |
 Do not use any method | 16 (32%) | ANN (4 times) | DT, RF, SVM (3 times, each) |