I'm trying to find the best eps values for DBSCAN. I set min_samples to 24 as I have 2 * 12 features passed through in x_set but I get the following error:
ValueError: Number of labels is 1. Valid values are 2 to n_samples - 1 (inclusive)
I know that the silhouette_score requires more than 1 cluster labels. Which might be causing this error based on similar error.
How do I solve this issue?
x_setcontains the following data:
x_set = [6.67933536e+00 1.65097236e+00 1.24573705e+00 1.01693195e+00
9.28128921e-01 7.82497904e-01 5.98319768e-01 1.13439548e-01
9.05382510e-04 5.42710767e-04 2.87522799e-04 1.90924073e-04]
def best_eps(x_set):
X = x_set
range_eps = [0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9]
for i in range_eps:
print("eps value is: "+str(i))
db = DBSCAN(eps=i, min_samples=24, metric='euclidean').fit(X)
core_samples_mask = np.zeros_like(db.labels_, dtype=bool)
core_samples_mask[db.core_sample_indices_] = True
labels = db.labels_
print(set(labels))
silhouette_avg = silhouette_score(X, labels)
print("For eps value ="+str(i), labels, "The average silhouette_score is :", silhouette_avg)
return
Choose parameters so that you get more than one cluster.
Also, Silhouette is not sensible to use on clusterings with Noise.
It will treat noise as a cluster, and assign it a very poor Silhouette.