FPA clust: Flower Pollination Algorithm for data clustering

FPA clust extracts the optimal cluster centers from training samples. The extracted cluster centers are then validated on test samples.

The two Matlab files, namely, Data_Clustering_FPA.m and Flower_Pollination.m are used to perform data clustering using Flower Pollination Algorithm.

The proposed partitional clustering approach extracts information in the form of optimal cluster centers from training samples. The extracted cluster centers are then validated on test samples.

Data_Clustering_FPA.m is the file which needs to be executed. This generates a synthetic dataset with predefined mean and standard deviation. Data is divided into train and test with predefined ratio. The users can vary these parameters. Training data is used to extract the cluster centers. Post the training phase, clustering is carried out on test dataset and results are displayed.

Flower_Pollination.m is the file called from Data_Clustering_FPA.m for extracting the optimal cluster centers from training dataset using Flower Pollination Algorithm. The file takes in training dataset with the upper & lower limits for each attribute as the input to the algorithm. The file returns the optimal cluster center to the Data_Clustering_FPA.m

In this illustration, a synthetic data with two class and two attributes is employed. The users can use their own custom datasets by replacing this synthetic data generation code using their own datasets. The users can vary FPA parameters based on the datasets they use for clustering.

The result metrics (optimal cluster centers, confusion matrix and classification error percentage) of clustering is displayed on the terminal. Further train, test data, initial agents of flower pollination and subsequent movement of agents to optimize are also visualized.