From: Deep learning in finance and banking: A literature review and classification
Identified F&B domains | Comparisons and ranking of commonly-used DL models |
---|---|
Exchange rate Prediction | DBN > FNN |
DBN + RBM > RW, ARMA, FNN | |
FNN + chaos theory + multi-objective evolutionary algorithms > FNN | |
CNN > MA | |
Prediction of stock market | Generative algorithm + FNN > FNN, SVM |
RNN + Stock2vec > LSTM, FNN | |
CNN > DT, SVM | |
CNN + special order feature set > LR > CNN | |
CNN > LSTM > RNN, ridge regression, Lasso, elastic net, random forecast, SVR, AdaBoost, gradient boosting, FNN, SVM, kNN, VAR, p-RBM. | |
LSTM + GARCH-type model > GARCH, dxponential weighted MA, LSTM | |
FNN + autoencoder + RBM > FNN, extreme learning machine, radial basis FNN | |
FNN + PCA > RNN, radial basis FNN | |
Stock trading | FNN > buy-and-hold method |
FNN > ARMA-GARCH | |
Fuzzy learning + RNN > CNN + RNN, LSTM + wavelet transforms + LSTM + SAEs > LSTM, RNN | |
RL + Q-learning > buy-and-hold method | |
RL + SARSA > buy-and-hold method | |
GA + RNN+ RL > RL | |
RL + DNN + transfer learning > RL, DNN | |
Portfolio management, | ListNet > RankNet |
RNN + RL > RNN | |
Banking default risk and credit | FNN + NLP > NLP |
CNN + feature selection > LR | |
LSTM + Event2vec > random forest+ Event2vec | |
Macroeconomic prediction | FNN > DT, LR |
FNN > SVM, random forest, classification tree | |
MLP (FNN) is the most widely used model in price prediction on oil. |