def main():
# Defining an object for the class MyKmeans
km = MyKmeans()
# Parsing the data file and acquiring a dataframe
parsedData = km.readData('digits-embedding.csv')
# Defining the numbers you need within the datafile to create a subset
sub_nums = [2, 4, 6, 7]
# Creating the subset
subset = createSubset(parsedData, sub_nums)
#--- The Tasks ---#
# (1) Visualizing the images based on the 2D features
visualizeData(subset, sub_nums)
# (2) Cluster the data with different values of K ∈ [2,4,8,16]. For each K
# repeat the experiment for 5 different times each with random centroids and
# calculate the average Silhouette Coefficient (SC) of each K after the 5 trials
K = [2, 4, 8, 16]
# >> K = 2
SC_avg_K2 = iterClustering(subset, parsedData, 5, 50, K[0])
# >> K = 4
SC_avg_K4 = iterClustering(subset, parsedData, 5, 50, K[1])
# # >> K = 8
SC_avg_K8 = iterClustering(subset, parsedData, 5, 50, K[2])
# # >> K = 16
SC_avg_K16 = iterClustering(subset, parsedData, 5, 50, K[3])
# Plotting results
plotAvgSC(K, [SC_avg_K2, SC_avg_K4, SC_avg_K8, SC_avg_K16])
def calAvgSC(clusters, parsedData):
km = MyKmeans()
# Intialize SC value
SC = 0
# Looping to obtain average SC value for all trials
for x in range(clusters.shape[0]):
temp = km.calculateSC(clusters[x], parsedData)
SC += temp
SC_avg = SC / clusters.shape[0]
return SC_avg
def iterClustering(subset, parsedData, iterNum, iterCount, K):
# Creating local object
km = MyKmeans()
cluster_K = [] # Initialize an empty array
for x in range(iterNum):
temp = km.cluster(subset, iterCount, K, [])
cluster_K.append(temp)
cluster_K = np.array(cluster_K)
SC_avg_k = calAvgSC(cluster_K, parsedData)
return SC_avg_k
def main():
# Defining an object for the class MyKmeans
km = MyKmeans()
# Parsing the data file and acquiring a dataframe
parsedData = km.readData('digits-embedding.csv')
# Defining the numbers you need within the datafile to create a subset
sub_nums = [0, 1, 2, 3, 4, 5, 6, 7, 8, 9]
# Creating the subset
subset = createSubset(parsedData, sub_nums)
#--- The Tasks ---#
# (1) Visualizing the images based on the 2D features
visualizeData(subset, sub_nums)