الگوریتم CLARANS

				
					
from pyclustering.cluster.clarans import clarans; 
from pyclustering.utils import timedcall; 
from sklearn import datasets 
#import iris dataset from sklearn library 
iris = datasets.load_iris(); 
#get the iris data. It has 4 features, 3 classes and 150 data points. 
data = iris.data 
"""! 
The pyclustering library clarans implementation requires 
list of lists as its input dataset. 
Thus we convert the data from numpy array to list. 
""" 
data = data.tolist() 
#get a glimpse of dataset 
print("A peek into the dataset : ",data[:4]) 
"""! 
@brief Constructor of clustering algorithm CLARANS. 

@details The higher the value of maxneighbor, the closer is CLARANS to K-Medoids, and the longer is each search of a local minima. 

@param[in] data: Input data that is presented as list of points (objects), each point should be represented by list or tuple. 
@param[in] number_clusters: amount of clusters that should be allocated. 

@param[in] numlocal: the number of local minima obtained (amount of iterations for solving the problem). 

@param[in] maxneighbor: the maximum number of neighbors examined.  

""" 

clarans_instance = clarans(data, 3, 6, 4); 

#calls the clarans method 'process' to implement the algortihm 

(ticks, result) = timedcall(clarans_instance.process); 

print("Execution time : ", ticks, "\n"); 


#returns the clusters  

clusters = clarans_instance.get_clusters(); 

#returns the mediods  

medoids = clarans_instance.get_medoids(); 

print("Index of the points that are in a cluster : ",clusters) 
print("The target class of each datapoint : ",iris.target) 

print("The index of medoids that algorithm found to be best : "medoids)