Skip to content
GitLab
Commit c1219a39 authored by Siu, Pui Chung's avatar Siu, Pui Chung
Browse files

Upload New File

parent c6204ff9
Hide whitespace changes
Inline Side-by-side
clustering: kmean clustering/assignment_kmeans_skeleton_python3.py 0 → 100644
View file @ c1219a39
#!/usr/bin/env python3
"""
Author: Jacky Siu Pui Chung
Student number = 1047527
Description: Implementation of the k-means clustering algorithm
Hints:
- write a function to obtain Euclidean distance between two points.
- write a function to initialize centroids by randomly selecting points
from the initial set of points. You can use the random.sample() method
- write a function to find the closest centroid to each point and assign
the points to the clusters.
- write a function to calculate the centroids given the clusters
- write a function to implement k-means
- write a function to calculate WGSS given the clusters and the points
- write a function to calculate BGSS given the clusters and the points
"""
# import statements
import random
import math
import copy
from statistics import mean
from statistics import stdev
def csv_parser(lines):
"""Return list of point coordinates as [[x1,y1,z1,...],[x2,y2,z2,...]]
lines: open file or list of lines. Expected format:
The file has a single header line.
Each line contains the coordinates for one data point, starting
with a label. A data point can be specified in arbitrary dimensions.
Output: List of lists with the coordinates of the points.
This function does not capture the labels of the data points. In case
they are needed later, this function should be adjusted.
"""
data_points = []
for line in lines:
items = line.strip().split(",")
try: #will fail on header line in file
data_points.append(list(map(float, items[1:]))) #skip label
except ValueError: #must be the header
continue
return data_points
def findeuclideandistance(pt1, pt2):
"""
Description: calculate euclidean distance between two data points
Input: pt1, list of a data point, length equal its dimension
pt2, list of a data point, length equal its dimension
Output: d, int of euclidean distance between pt1 and pt2
"""
d = math.sqrt(sum([(a - b) ** 2 for a, b in zip(pt1, pt2)]))
return d
def centroidinit(pointslist, k):
"""
Description: initialize centroid by randomly sampling from the dataset
Input: pointslist, list of list containing data points with same number
of dimensions
k, int of number of clusters
Output: init_centroid_list, list of list containing data points of the
initialized centroids
clusters, dictionary with the key as 0, 1, 2,..... and values
as the respective initialized centroid
"""
init_centroid_list = random.sample(pointslist, k)
clusters = {clusternum:[init_centroid] for clusternum, init_centroid in \
enumerate(init_centroid_list)}
return init_centroid_list, clusters
def findclosestcentroid(pointslist, clustersdict):
"""
Description: for each data point, find closest centroid
Input: pointslist, list of list containing data points with same number
of dimensions
clustersdict, dictionary with the key as 0, 1, 2,..... and values
as list of list, where list[0] are respective initialized centroid
for that cluster
Output: clustersdict, dictionary with keys as the cluster number and
values as a list of list containing all data points associated
with that cluster
labellist, list that is the same length as the pointslist, where
each point is associated with a cluster
"""
labellist = []
clusterslist = [init_centroid[0] for init_centroid in clustersdict.values()]
#removal of centroid from cluster list during initialization
pointslist_init = [datapoints for datapoints in pointslist if datapoints \
not in clusterslist]
for unassigned_points in pointslist_init:
dis = []
for i in range(len(clusterslist)):
dis.append(findeuclideandistance(clusterslist[i], unassigned_points))
clustersdict[dis.index(min(dis))].append(unassigned_points)
labellist.append(dis.index(min(dis)))
return clustersdict, labellist
def reassigncentroid(clustersdict):
"""
Description: reassign the centroid for each cluster by calculating the mean
between all data points within the cluster
Input: clustersdict, dictionary with keys as the cluster number and
values as a list of list containing all data points associated
with that cluster
Output:
clustersdict, clustersdict, dictionary with the key as cluster
number and values as list of list, where list[0]
are respective reassigned centroid for that cluster
"""
for cluster, data_points_in_cluster in list(clustersdict.items()):
avg = [mean(dimension) for dimension in zip(*data_points_in_cluster)]
clustersdict[cluster] = [avg]
return clustersdict
def checkconvergence(clustersdictbefore, clustersdictafter):
"""
Description: compare two clustersdict to check if they have the same
centroid at each cluster
Input: clustersdictbefore, dictionary with keys as the cluster number and
values as a list of list where list[0]
are respective reassigned centroid for that cluster
clustersdictafter, dictionary with keys as the cluster number and
values as a list of list where list[0]
are respective reassigned centroid for that cluster
Output:
True when they share same centroid at each cluster, false when
they do not
"""
for keys in clustersdictbefore.keys():
if clustersdictbefore[keys] != clustersdictafter[keys]:
return False
return True
def checkemptycluster(clustersdict):
"""
Description: check whether if a cluster has no data points associated
Input: clustersdict, dictionary with keys as the cluster number and
values as a list of list where list[0]
are respective reassigned centroid for that cluster
Output:
True when they share same centroid at each cluster, false when
they do not
"""
for keys in clustersdict.keys():
#cluster only contain centroid
if len(clustersdict[keys]) == 1:
return True
return False
def kmeansclusteringwrapper(pointslist, k):
"""
Description: wrapper that connect the k mean clustering algorithm
Input: pointslist, list of list containing data points with same number
of dimensions
k, int of number of clusters
Output:
finallabellist, list that associate each point with a cluster
clustersdictfinal, dictionary with keys as the cluster number and
values as a list of list containing all data points associated
with that cluster
numtoconvergence, int of number of times it take to converge
"""
# number of times it takes to converge, minimum once
numtoconvergence = 1
init_centroid_list, clustersdict = centroidinit(pointslist, k)
#deep copy
clusersdictafter = copy.deepcopy(clustersdict)
clustersdictafter, updatedlabellist= findclosestcentroid(pointslist,\
clusersdictafter)
#if centroid has no data points associated, reinitiate centroid
if checkemptycluster(clustersdictafter):
kmeansclusteringwrapper(pointslist, k)
while checkconvergence(clustersdict, clustersdictafter) == False:
numtoconvergence += 1
clustersdict = copy.deepcopy(clustersdictafter)
clustersdictafter = copy.deepcopy(clustersdictafter)
clustersdictafter, updatedlabellist= findclosestcentroid(pointslist, \
clustersdictafter)
clustersdictafter = reassigncentroid(clustersdictafter)
#when converges
clustersdictfinal, finallabellist = findclosestcentroid(pointslist, \
clustersdictafter)
if checkemptycluster(clustersdictafter):
kmeansclusteringwrapper(pointslist, k)
return finallabellist, clustersdictfinal, numtoconvergence
def WGSS(clustersdict):
"""
Description: calculate WGSS for clusters
Input: clustersdict, dictionary with keys as the cluster number and
values as a list of list containing all data points associated
with that cluster
Output:
WGSS_total, int of the WGSS score of the cluster
"""
WGSS_total = 0
for keys in clustersdict.keys():
data_points = clustersdict[keys]
centroid = data_points[0]
WGSS = [findeuclideandistance(data_point, centroid)**2 for \
data_point in data_points]
WGSS = mean(WGSS)
WGSS_total += WGSS
return WGSS_total
def BGSS(clustersdict, pointlist):
"""
Description: calculate BGSS for clusters
Input: clustersdict, dictionary with keys as the cluster number and
values as a list of list containing all data points associated
with that cluster
pointlist:list of list containing data points with same number
of dimensions
Output:
BGSS_total, int of the WGSS score of the cluster
"""
BGSS_total = 0
mean_centroid = [mean(dimension) for dimension in zip(*pointlist)]
for key in clustersdict.keys():
centroid = clustersdict[key][0]
d = findeuclideandistance(centroid, mean_centroid)
n = len(clustersdict[key]) - 1
BGSS_total += d**2 * n
return BGSS_total
def clusterevaulation(clustersdict, pointlist):
"""
Description: wrapper that calculate W score for clusters
Input: clustersdict, dictionary with keys as the cluster number and
values as a list of list containing all data points associated
with that cluster
pointlist:list of list containing data points with same number
of dimensions
Output:
W, int of the W score of the cluster
"""
clusters_for_WGSS = copy.deepcopy(clustersdict)
clusters_for_BGSS = copy.deepcopy(clustersdict)
WGSS_score = WGSS(clusters_for_WGSS)
BGSS_score = BGSS(clusters_for_BGSS, pointlist)
W = WGSS_score / BGSS_score
return W
def statistical_calc(pointlist, k, repetition):
"""
Description: wrapper that calculate W score, niter and final labels
for clusters in a given repitition
Input: pointlist:list of list containing data points with same number
of dimensions
k, int of number of clusters
repitition: number of times to perform k mean clustering
on the pointlist with given k
Output:
W_list, list of int of the W score of the cluster
numtoconvergence_list: list of int of the niter of each k mean
clustering
finallabellist: list of list of the label for each point
in the pointlist
"""
W_list = []
numtoconvergence_list = []
finallabellist = []
while len(W_list) < repetition:
finallabels, clustersdictfinal, numtoconvergence = \
(kmeansclusteringwrapper(pointlist, k))
W = clusterevaulation(clustersdictfinal, pointlist)
W_list.append(W)
numtoconvergence_list.append(numtoconvergence)
finallabellist.append(finallabels)
return W_list, numtoconvergence_list, finallabellist
if __name__ == "__main__":
data_points_2dtest_2021 = csv_parser(open("datasets/2dtest_2021.csv"))
# the code below should produce the results necessary to answer
# the questions. In other words, if we run your code, we should see
# the data that you used to answer the questions.
W_list_2dtest_2021, numtoconvergence_list_2dtest_2021,\
finallabellist_2dtest_2021 = statistical_calc(data_points_2dtest_2021, 3, 20)
print("1a)", mean(numtoconvergence_list_2dtest_2021), \
stdev(numtoconvergence_list_2dtest_2021))
print("1b) lowest W = ", min(W_list_2dtest_2021), "label list = ", \
finallabellist_2dtest_2021[\
W_list_2dtest_2021.index(min(W_list_2dtest_2021))])
print("1c)", mean(W_list_2dtest_2021), stdev(W_list_2dtest_2021))
print("1d)", min(W_list_2dtest_2021))
k_list = [3, 4, 5, 6]
data_points_LargeSet_1_2021 = csv_parser(open("datasets/LargeSet_1_2021.csv"))
LargeSet_1_2021_W_list = []
LargeSet_1_2021_ntc_list = []
for k in k_list:
W_list , numtoconvergence_list, finallabellist= \
statistical_calc(data_points_2dtest_2021, k, 50)
LargeSet_1_2021_W_list.append((mean(W_list), stdev(W_list)))
LargeSet_1_2021_ntc_list.append(\
(mean(numtoconvergence_list), stdev(numtoconvergence_list)))
print("2a)")
for i in range(len(LargeSet_1_2021_ntc_list)):
print("For niter, k = ", i + 3, ": mean = ", LargeSet_1_2021_ntc_list[i][0],\
"sd = ", LargeSet_1_2021_ntc_list[i][1])
print("2b)")
for i in range(len(LargeSet_1_2021_W_list)):
print("For W, k = ", i + 3, ": mean = ", LargeSet_1_2021_W_list[i][0],\
"sd = ", LargeSet_1_2021_W_list[i][1])
print("2c)")
for k in k_list:
W_list , numtoconvergence_list, finallabellist= statistical_calc\
(data_points_2dtest_2021, k, 50)
print("for k = {0}".format(k))
print("smallest W value = ", min(W_list))
print(finallabellist[W_list.index(min(W_list))])
print("2d) The lowest W value was selected because it means it has the\
strongest cluster structure")
print("2e) Selecting the lowest W value is better as not all convergence \
lead to the global minimum (or at least lowest known)")
print("3) K mean clustering assume that the variance of \
the distribution within a cluster is spherical,\
which is not applicable in the dataset")
print("4) For each value of k, a range of repitition can \
be used for cross validation to find the lowest mean W")
print("5) with an outlier in a cluster, it dramatically shifts \
the reassignment of the centroid as the mean\
is sensitive to the outlier's position")
print("6) the k mean clustering algorithm is not deterministic,\
as the initialization of the centroid are \
randomized")
print("7) O(N(D+K)), N being the number of data points \
in the dataset, D being the dimension, and K \
being the number of clusters")
print("8) the time complexity = O(NKI), N being the number \
of data points in the dataset, K being \
the number of clusters, and I being the number of iterations til convergence.")
Supports Markdown
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment