Advances, Systems and Applications
Prediction Methods | ||
---|---|---|
Method | Advantage | Disadvantage |
SVM | Possessing good generalization ability and being insensitive to noisy data [44]. | difficulty occurs in selecting suitable kernel [45]. |
KNN | The principle is simple and there are few influencing factors [46]. | The computational complexity is high, and the algorithm performance is easily affected by the samples [17]. |
ANN | Able to learn non-linear features of data [47]. | Easily trapped in local optima [48]. |
LSTM | Has the ability to capture long-term sequence features [49]. | difficult to balance the depth and operation time complexity of its model [50]. |
CNN-LSTM | Reducing the risk of overfitting can uncover more feature information [51, 52]. | There is a problem of neural network model degradation [53]. |
GCN | High efficiency, possessing the ability to effectively learn data relationships, and providing interpretability [54]. | There is a problem of excessive smoothing [55]. |