New clustering method uncovers hidden regularities in data
One dataset, one model? This approach does not always produce the most useful insights, as a dataset often contains many different relationships. To better understand them, researchers at the Paluno Research Institute in the Faculty of Computer Science at the University of Duisburg-Essen have developed a method that clusters data according to mathematical functions. The key feature is that…
Researchers at the Paluno Research Institute have devised a novel method to uncover hidden regularities in data, a technique known as CluBS (Clustering Behavioral Similarity). This innovative approach identifies groups of data points that can be accurately described by the same mathematical function, rather than relying on proximity in feature space.
By using symbolic regression to discover the underlying functions and iteratively assigning data points to these groups, CluBS can reveal complex behaviors present in both real-world and synthetic datasets. The method proved particularly effective in identifying multiple functions that better described distinct behaviors than a single model.
The researchers plan to further develop CluBS to enhance its predictive capabilities, specifically aiming to enable the model to select the most appropriate function without prior knowledge of target values.
Written by urgent.news from Phys.org's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
