Many of the machine learning algorithms use k-means clustering to improve the clasification performance based on human physiological signals.
However, while some studies uses a subject specific clustering (where subjects are divided into subgroups such that each subject is finally assigned to a specific subgroup), some other studies used epoch based clustering (where multiple epochs from the same subject's data are assigned to multiple subgroups. This is also extended to multiple subjects).
Why are these two approaches required and what could be the advantages/disadvantafges of one over another?
From what I understand, for each subject in your data, you have a sequence of observations over epochs. Based on this assumption, my understanding is that the subject-specific approach clusters the subjects based on the average behavior of the subject, while the epoch-specific approach clusters the individual observations. The former would tell you about the average nature of the subjects while the latter can give you an idea of how the behavior of the subjects change over epochs.
- Badges
- Science topic