In bsoid_app/extract_features.py it determines the number of dimensions by find what number of components explain 70% of the variance when using PCA. However, I don't see why this would be meaningful for UMAP when UMAP and PCA are different algorithms.
Perhaps this step and the clustering step should be put together: Test out a range of different number of components and then cluster the data and see which number of components will give the best clustering. This may take a long time, however.
In bsoid_app/extract_features.py it determines the number of dimensions by find what number of components explain 70% of the variance when using PCA. However, I don't see why this would be meaningful for UMAP when UMAP and PCA are different algorithms.
Perhaps this step and the clustering step should be put together: Test out a range of different number of components and then cluster the data and see which number of components will give the best clustering. This may take a long time, however.