I’m working on a machine learning problem and trying to boost the effectiveness of the model. I have use Randomized search to find the best hyper-parameters for my model and it increased the score but not to a satisfactory level.
I decided to employ pca(Principal Component Analysis).

from sklearn.decomposition import PCA pca = PCA(n_composition=0.95)

I scaled and transformed my data using one-hot and standard scaler for the neccessary preprocessing.
After including PCA to my model it further reduced the score, would LDA be neccessay for this problem?
and is there a more efficient way to actually use PCA other than the above to better my output?
The models are XGBoost and LightGBM.
My machine learning problem is a binary classification, note that my initial model isn’t overfitting

I tried PCA and waiting to try LDA, I was expecting to increase my score.