From: Chronic kidney disease prediction using machine learning techniques
No. | Author | Technique applied | Claimed outcome | Draw back |
---|---|---|---|---|
1 | Salekin and Stankovic [9] | K-NN, RF, and NN, Wrapper approach and Embedded approach | Detection F1-score of RF 99.8 | Small dataset size with missing values was used; Severity level prediction was not included |
2 | Tekale et al. [10] | DT and SVM | DT and SVM with an accuracy of 91.75 and 96.75 respectively | Dataset size need to increased, Severity level prediction was not included. Only to classifiers result compared |
3 | Priyanka et al. [12] | NB, KNN, SVM, DT, and ANN. NB | NB, KNN, SVM, DT, and ANN. NB accuracy is 94.6% | Small size dataset. No stages prediction. Feature extraction was not carried out and classification accuracy needs improvement |
4 | Yashfi [14] | RF and ANN | RF and ANN with an accuracy of 97.12% and 94.5% | Small size dataset and no stages prediction |
5 | Rady and Anwar [15] | PNN, MLP, SVM, RBF | PNN, MLP, SVM, RBF accuracy of 96.7, 60.7, 87, and 51.5 respectively | Small data size. the algorithms, which is not appropriate for small dataset size were used |
6 | Alsuhibany et al. [16] | EDL-CDSS technique, ADASYN based outlier detection and QOBOA based hyperparameter tuning | The proposed EDL-CDSS method has depicted the other approaches with the superior sensy and specy of 0.9680 and 0. 9702 compared to ACO, FNC, KELM, CNN-GRU, DBN, DT, MLP and D-ACO | The size of benchmark dataset is small. IoT and benchmark data integration is not clearly explained. The study focus only on binary classification (ckd or not-ckd) |
7 | Poonia et al. [17] | KNN, ANN, SVM, NB and LR with Chi-Square and RFE | KNN, ANN, SVM, NB, LR accuracy of 66.25%,65%,97.5%,95%,97.5% respectively | Small size dataset and no stages prediction. Accuracy needs improvement |