You should automate this whole discovery process to try different clustering algorithms with different hyperparameter settings. The following code will show you a simple way of doing that:
# You will create a list of algorithms to testfrom sklearn.cluster import MeanShift, estimate_bandwidth, SpectralClusteringfrom hdbscan import HDBSCAN# bandwidth estimate for MeanShift algorithm to work properlybandwidth = estimate_bandwidth(X, quantile=0.3, n_samples=100)estimators = [{'estimator': KMeans, 'args': (), 'kwargs': {'n_clusters': 5}}, {'estimator': DBSCAN, 'args': (), 'kwargs': {'eps': 0.5}}, {'estimator': AgglomerativeClustering, 'args': (), 'kwargs': {'n_clusters': 5, 'linkage': 'ward'}}, {'estimator' ...
