A New Approach of Data Clustering Using a Flock of Agents

Abstract

This paper presents a new bio-inspired algorithm (FClust) that dynamically creates and visualizes groups of data. This algorithm uses the concepts of a flock of agents that move together in a complex manner with simple local rules. Each agent represents one data. The agents move together in a 2D environment with the aim of creating homogeneous groups of data. These groups are visualized in real time, and help the domain expert to understand the underlying structure of the data set, like for example a realistic number of classes, clusters of similar data, isolated data. We also present several extensions of this algorithm, which reduce its computational cost, and make use of a 3D display. This algorithm is then tested on artificial and real-world data, and a heuristic algorithm is used to evaluate the relevance of the obtained partitioning.

Swarm intelligence for clustering dynamic data sets for web usage mining and personalization.

“…More recently, an approach based on flocks of agents, known as FClust, was used for data clustering [129,128,117,116]. The flock-based approach holds a great promise, and being the most recent of all the swarm-based clustering approaches, it has been the least studied.…”

“…Self-organizing maps (SOM) are also a popular way for data visualization and some of its applications include the visualization of web-based social networks [119,143]. Lately, swarm intelligence has also been used for data visualization [128,129,117].…”

“…Lately, in [117], a detailed flock clustering algorithm (FClust) was presented with a stopping criterion and automated cluster extraction algorithm. An application of this approach to Web usage mining can be found in [128].…”

“…[117] Input: Dataset. Output: Visualization of interaction between the data items. Agents corresponding to more similar items are located closer in the 2D visualization panel.…”

“…However, if the amplitude is too high, the agents may move further than the desired location. The minimum amplitude is empirically set to t x d th [117]. At the end of each iteration, the updated agent coordinates are computed, and all the agents are moved to their updated positions simultaneously (lines 28 to 31).…”