-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathk-means.py
183 lines (139 loc) · 8.33 KB
/
k-means.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
from utils import CSVMemoryEfficientOperations, jaccard_similarity, cosine_similarity
import random, time
def hierarchicalClustering(data:list, options:list, k:int):
clusters = [{"centroid": row, "members" : [{"distance": 0, "payload": row}]} for row in data]
for n in range(0, len(clusters) - k):
closest_distance = float('inf')
cluster_i = float('inf')
cluster_j = float('inf')
for i in range(0, len(clusters)):
for j in range(0, len(clusters)):
if i != j:
distance = 0
for option in options:
feature_value_1 = clusters[i]["centroid"][option["feature"]]
feature_value_2 = clusters[j]["centroid"][option["feature"]]
distance_f = option["distance_f"]
distance = distance + distance_f(feature_value_1, feature_value_2)*option["weight"]
if distance < closest_distance and distance != 0:
closest_distance = distance
cluster_i = clusters[i]
cluster_j = clusters[j]
cluster_merged = {"centroid": None, "members": cluster_i["members"] + cluster_j["members"]}
cluster_merged["centroid"] = findCentroid(cluster_merged["members"], options)
clusters.remove(cluster_i)
clusters.remove(cluster_j)
clusters.append(cluster_merged)
return clusters
def findCentroid(members:list, options:list):
curr_distance = float('inf')
centroid = 0
for i in range(0, len(members)):
distance = 0
for j in range(0, len(members)):
if i != j:
for option in options:
feature_value_1 = members[i]["payload"][option["feature"]]
feature_value_2 = members[j]["payload"][option["feature"]]
distance_f = option["distance_f"]
distance = distance + (distance_f(feature_value_1, feature_value_2)*option["weight"])**2
if distance < curr_distance:
curr_distance = distance
centroid = members[i]
for i in range(0, len(members)):
for option in options:
feature_value_1 = centroid["payload"][option["feature"]]
feature_value_2 = members[j]["payload"][option["feature"]]
distance_f = option["distance_f"]
members[i]["distance"] = distance_f(feature_value_1, feature_value_2)
members.sort(key=lambda x : x["distance"])
return centroid["payload"]
def findRepresentatives(clusters):
representatives = []
for i in range(0, len(clusters)):
lower_bound = clusters[i]["members"][int(len(clusters[i]["members"])*(60/100))]["distance"]
upper_bound = clusters[i]["members"][int(len(clusters[i]["members"])*(80/100))]["distance"]
representatives_i = []
for j in range(0, len(clusters[i]["members"])):
if clusters[i]["members"][j]["distance"] >= lower_bound and clusters[i]["members"][j]["distance"] <= upper_bound:
representatives_i.append(clusters[i]["members"][j])
if len(representatives_i) > 7:
representatives_i = random.sample(representatives_i, 7)
clusters[i]["members"] = []
representatives.append(representatives_i)
return representatives
def associateToCluster(row:list, representatives:list, options:list):
curr_distance = float('inf')
curr_i = 0
for i in range(0, len(representatives)):
for j in range(0, len(representatives[i])):
distance = 0
for option in options:
feature_value_1 = row[option["feature"]]
feature_value_2 = representatives[i][j]["payload"][option["feature"]]
distance_f = option["distance_f"]
distance = distance + distance_f(feature_value_1, feature_value_2)*option["weight"]
if distance < curr_distance:
curr_distance = distance
curr_i = i
return [curr_i, curr_distance] + row
def CUREDiskBased(in_file:str, process_id:str, options:list, k:int, chunk_size:int):
offsets = CSVMemoryEfficientOperations.getLinesOffsets(in_file, chunk_size)
headings = CSVMemoryEfficientOperations.getHeadings(in_file)
initial_load = CSVMemoryEfficientOperations.readChunk(in_file, 0, offsets, chunk_size, options)
clusters = hierarchicalClustering(initial_load, options, k)
representatives = findRepresentatives(clusters)
chunk = initial_load
del initial_load
count = 0
while True:
rows_output = []
for i in range(0, len(chunk)):
line = associateToCluster(chunk[i], representatives, options)
for option in options:
lambda_f = option["encode"] if option["encode"] else lambda x: x
line[option["feature"] + 2] = lambda_f(line[option["feature"] + 2])
rows_output.append(line)
CSVMemoryEfficientOperations.writeChunk(process_id + ".csv", count, rows_output, ["cluster", "distance"] + headings)
count = count + 1
chunk = CSVMemoryEfficientOperations.readChunk(in_file, count, offsets, chunk_size, options)
if chunk == None or len(chunk) < 1:
break
CSVMemoryEfficientOperations.sortByKey(process_id + ".csv", process_id + "_sorted.csv", 0)
CSVMemoryEfficientOperations.sortByKeyByGroup(process_id + "_sorted.csv", process_id + ".csv", 1, 0)
CSVMemoryEfficientOperations.deleteFile(process_id + "_sorted.csv")
representatives_output = []
for i in range(0, len(representatives)):
for j in range(0, len(representatives[i])):
line = representatives[i][j]["payload"]
for option in options:
lambda_f = option["encode"] if option["encode"] else lambda x: x
line[option["feature"]] = lambda_f(line[option["feature"]])
representatives_output.append([i] + [representatives[i][j]["distance"]] + line)
CSVMemoryEfficientOperations.writeChunk(process_id + "_representatives.csv", 0, representatives_output, ["cluster", "distance"] + headings)
CSVMemoryEfficientOperations.sortByKey(process_id + "_representatives.csv", process_id + "_representatives_sorted.csv", 0)
CSVMemoryEfficientOperations.sortByKeyByGroup(process_id + "_representatives_sorted.csv", process_id + "_representatives.csv", 1, 0)
CSVMemoryEfficientOperations.deleteFile(process_id + "_representatives_sorted.csv")
centroids_output = []
for i in range(0, len(clusters)):
line = clusters[i]["centroid"]
centroids_output.append([i] + line)
CSVMemoryEfficientOperations.writeChunk(process_id + "_centroids.csv", 0, centroids_output, ["cluster"] + headings)
CSVMemoryEfficientOperations.sortByKey(process_id + "_centroids.csv", process_id + "_centroids_sorted.csv", 0)
CSVMemoryEfficientOperations.deleteFile(process_id + "_centroids_sorted.csv")
return
start_exec = time.perf_counter()
#a) d1: jaccard similarity based on the genres of the movies
d1 = [{"feature": 2, "decode": lambda x: x.split("|"), "encode": lambda x: "|".join(x), "weight": 1, "distance_f": jaccard_similarity}]
CUREDiskBased("movies_joined_cleaned.csv", "exercise1_a", d1, 8, 650)
#b) d2: jaccard similarity based on the tags of the movies
d2 = [{"feature": 4, "decode": lambda x: x.split("|"), "encode": lambda x: "|".join(x), "weight": 1, "distance_f": jaccard_similarity}]
CUREDiskBased("movies_joined_cleaned.csv", "exercise1_b", d2, 8, 650)
#c) d3: cosine similarity based on the ratings of the movies
d3 = [{"feature": 3, "decode": lambda x: x.split("|"), "encode": lambda x: "|".join(x), "weight": 1, "distance_f": cosine_similarity}]
CUREDiskBased("movies_joined_cleaned.csv", "exercise1_c", d3, 8, 650)
#d) d4 = 0.3*d1 + 0.25*d2 + 0.45*d3
d4 = [{"feature": 2, "decode": lambda x: x.split("|"), "encode": lambda x: "|".join(x), "weight": 0.3, "distance_f": jaccard_similarity}, {"feature": 3, "decode": lambda x: x.split("|"), "encode": lambda x: "|".join(x), "weight": 0.45, "distance_f": cosine_similarity}, {"feature": 4, "decode": lambda x: x.split("|"), "encode": lambda x: "|".join(x), "weight": 0.25, "distance_f": jaccard_similarity}]
CUREDiskBased("movies_joined_cleaned.csv", "exercise1_d", d4, 8, 650)
stop_exec = time.perf_counter()
print("TOTAL EXEC TIME", f"{stop_exec - start_exec:0.4f} seconds")