If I want to generate a synthesized dataset to be used in testing my clustering algorithm, what should I consider while preparation? I have found a python method called make_blobs. However, I don't know how to make the boundaries of its features to be different.
Example:
Age boundaries: 20-40
Sex boundaries: 0,1
etc...
Can you briefly explain how make_blobs works?
Generally speaking, the most important point to consider while designing artificial data for clustering is to ensure that there actually exist some cluster structure in the generated data.
Agree with Shounak Datta . What you can do is define the distribution of features for each cluster and generate the data its samples randomly (if distributions are different that should imply there is a cluster structure in the data).
High-Variability and Self-Similarity are two parameters that we should we care about when generating synthesized dataset.
- Badges
- Science topic