Embedded Files
Unlike Experiment 1, feature selection based on domains is not applied. Moreover, due to the high accuracy of GDS-15 in Experiment 1, we focused on MMSE for Experiment 2 by applying different techniques to increase the accuracy.
Unlike Experiment 1, feature selection based on domains is not applied. Moreover, due to the high accuracy of GDS-15 in Experiment 1, we focused on MMSE for Experiment 2 by applying different techniques to increase the accuracy.
Before feature selection, train-test split (80%,20%) has been conducted to remain test set unseen.
Before feature selection, train-test split (80%,20%) has been conducted to remain test set unseen.
We focused on SelectKBest feature selection technique in this experiment:
SelectKBest
We choose the features that score >30, which are the top 9 features.
Visualization of the features
The final train and test set, X_train_f and X_test_f only consists of the top 9 features
Page updated
Report abuse