Why does dimension reduction reduce accuracy?

Question

I used python sklearn kit and used PCA/ISOMAP/TruncatedSVD

    from sklearn.manifold import Isomap
    isomap = Isomap(n_components=8)
    train_feature_x = i