-
Notifications
You must be signed in to change notification settings - Fork 0
/
shantanu_deshmukh_knn.py
309 lines (279 loc) · 12.4 KB
/
shantanu_deshmukh_knn.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
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
'''
Created on Sep 7, 2017
@author: Shantanu Deshmukh
'''
from matplotlib import pyplot as plt
import sys,os
class Agent(object):
'''
Agent class senses input, perform computations on input(function()) and takes action(actuator())
'''
def __init__(self, k, traRows, attributeInfo, optimizationParams):
'''
initializes k, training percept, information of each attribute and optimization parameters like data distribution, distance algorithm
'''
self.k=k
self.traRows=traRows
self.attributeInfo = attributeInfo
self.optimizationParams = optimizationParams
def sensor(self, tstRow):
'''
senses the testing row and passes to the function
'''
return self.function(tstRow)
def function(self, tstRow):
'''
Finds k nearest neighbours of the test row, according to the optimization parameters, and passes to the actuator to find the label
'''
neighbourDistance={}
neighbours=[]
if self.optimizationParams["dataDistribution"] == 1:
tstRow = generalize(tstRow,self.attributeInfo)
for i in range(len(self.traRows)):
if self.optimizationParams["dataDistribution"] == 1:
'''
Enters here if user has selected to distribute the data and as per the distance algorithm fill in the neighbourDistance list
'''
if self.optimizationParams["distanceAlgo"] == 1:
attrType = []
for info in self.attributeInfo:
if info["maxRange"]-info["minRange"]<4:
attrType.append("cat")
else:
attrType.append("conti")
neighbourDistance[i]=intelligentDistance(tstRow, generalize(self.traRows[i], self.attributeInfo), attrType)
elif self.optimizationParams["distanceAlgo"] == 2:
neighbourDistance[i]=manhattanDistance(tstRow, generalize(self.traRows[i], self.attributeInfo))
else:
neighbourDistance[i]=euclidianDistance(tstRow, generalize(self.traRows[i], self.attributeInfo))
else:
'''
if user does not want to distribute data evenly, calculate distance without generalize() and fill the neighboutDistance list
'''
if self.optimizationParams["distanceAlgo"] == 1:
attrType = []
for info in self.attributeInfo:
if info["maxRange"]-info["minRange"]<4:
attrType.append("cat")
else:
attrType.append("conti")
neighbourDistance[i]=intelligentDistance(tstRow, self.traRows[i],attrType)
elif self.optimizationParams["distanceAlgo"] == 2:
neighbourDistance[i]=manhattanDistance(tstRow, self.traRows[i])
else:
neighbourDistance[i]=euclidianDistance(tstRow, self.traRows[i])
for i in sorted(neighbourDistance, key=neighbourDistance.get):
neighbours.append(self.traRows[i])
return self.actuator(neighbours[0:int(self.k)])
def actuator(self, neighbours):
'''
From the nearest neighbours counts same type of neighours and returns label of neighbour with maximumm count
'''
labels={}
for neighbour in neighbours:
label = neighbour[len(neighbour)-1]
if labels.has_key(label):
labels[label]+=1
else:
labels[label]=1
maxCount = 0
predictedLabel = ""
for label in labels.items():
if label[1]>maxCount:
maxCount=label[1]
predictedLabel=label[0]
return predictedLabel
def generalize(row, attributeInfo):
'''
Generalizes attribute data in a row.
for e.g. if value of one attribute is 250 and the range of attribute is (150,350) then this function will reduce its range to (0,100) so its new value will be 50
This will be done on each attribute so that each attribute will have same weight while calculating distance
'''
generalizedRow = []
i=0
for info in attributeInfo:
generalizedRow.append(round(100/(info["maxRange"]-info["minRange"])*(float(row[i])-info["minRange"]),2))
i+=1
generalizedRow.append(row[i])
return generalizedRow
def euclidianDistance(row1,row2):
'''
Calculates euclidian distance between 2 rows
'''
distance = 0
for i in range(len(row1)-1):
distance += pow((float(row1[i]) - float(row2[i])), 2)
return distance
def manhattanDistance(row1,row2):
'''
Calculates manhattan distance between 2 rows
'''
distance = 0
for i in range(len(row1)-1):
distance += abs(float(row1[i]) - float(row2[i]))
return distance
def intelligentDistance(row1,row2,attrType):
'''
If an attribute is Categorical with less then 4 categories, it will use manhattan distance else it will use euclidian distance
'''
distance = 0
for i in range(len(row1)-1):
if(attrType[i] == "cat"):
distance += abs(float(row1[i]) - float(row2[i]))
else:
distance += pow((float(row1[i]) - float(row2[i])), 2)
return distance
class Environment(object):
'''
Environment for the agent. It feeds percepts to the agent's sensor function, calculates accuracy and plots graphs accordingly.
'''
def __init__(self, k, folderName, optimizationParams):
'''
Sets value of k, folderName and optimizationParams as set by the user during bigbang()
initializes accuracy and attributeInfo lists
'''
self.folderName = folderName
self.k = k
self.attributeInfo = []
self.accuracy = []
self.optimizationParams = optimizationParams
def createAgent(self, trainingRows):
'''
Creates agent with the essential parameters
'''
return Agent(self.k, trainingRows, self.attributeInfo, self.optimizationParams)
def classifyAndComputeAccuracy(self):
'''
Calls agent to get the test data classified and then computes accuracy
'''
isAttrbuiteInfoExtracted = 0;
for j in range(1,11):
with open(self.folderName+"/"+self.folderName+"-10dobscv-"+repr(j)+"tra.dat",'r') as dataFile:
traRows = []
i = 0
for line in dataFile:
line = line.strip()
if line[0] != '@':
traRows.insert(i, line.split(","))
i+=1
elif not isAttrbuiteInfoExtracted:
self.extractAttributeInfo(line.split())
if len(self.attributeInfo) > 0:
isAttrbuiteInfoExtracted = 1
agent = self.createAgent(traRows)
tstCount,correctPrediction = 0,0
with open(self.folderName+"/"+self.folderName+"-10dobscv-"+repr(j)+"tst.dat",'r') as tstRows:
for tstRow in tstRows:
tstRow = tstRow.strip()
if tstRow[0] != '@':
tstRow = tstRow.split(",")
pLabel = agent.sensor(tstRow)
if tstRow[len(tstRow)-1] == pLabel:
correctPrediction+=1
tstCount+=1
self.accuracy.insert(j, round(correctPrediction/float(tstCount),2))
def process(self):
'''
If k is 0, evaluates accuracy for different values of k. Otherwise, computes accuracy for each iteration of percept for given value of k.
Draws accuracy graphs accordingly
'''
if int(self.k) == 0:
averageAccuracy = {}
minAccuracy = 0
optimalK = 0
for kNew in range(1,self.optimizationParams["maxK"]):
self.attributeInfo = []
self.k = kNew
self.classifyAndComputeAccuracy()
averageAccuracy[kNew]=sum(self.accuracy)/len(self.accuracy)
for kNew in averageAccuracy.keys():
if averageAccuracy[kNew] > minAccuracy:
minAccuracy = averageAccuracy[kNew]
optimalK = kNew
x = range(1,self.optimizationParams["maxK"])
y = averageAccuracy.values()
title = "For k = "+repr(optimalK)+" accuracy is maximum ("+repr(round(averageAccuracy[optimalK],2))+")"
supTitle = "Accuracy vs k plot"
self.drawGraph(x, y, title, supTitle, "Value of k", "Accuracy")
else:
self.classifyAndComputeAccuracy()
overallAcc = sum(self.accuracy)/len(self.accuracy)
x = range(0,len(self.accuracy))
y = self.accuracy
title = "For k = "+repr(self.k)+", average accuracy obtained is "+repr(round(float(overallAcc),2))
supTitle = "Classification Accuracy Plot"
self.drawGraph(x, y, title, supTitle, "Iteration", "Accuracy")
def extractAttributeInfo(self,row):
'''
Reads attribute info at the start of the training data and intelligently extracts information
'''
if row[0] == "@attribute" and row[1].find("{")<0:
attribute={}
attribute["name"] = row[1]
attribute["minRange"] = float(row[2][row[2].find("[")+1:row[2].find(",")])
attribute["maxRange"] = float(row[2][row[2].find(",")+1:row[2].find("]")])
self.attributeInfo.append(attribute)
def drawGraph(self,x,y,title,supTitle,xlabel,ylabel):
'''
Plots graph as per the information provided in parameters
'''
plt.suptitle(supTitle)
plt.title(title)
plt.xlabel(xlabel)
plt.ylabel(ylabel)
plt.plot(x,y,'b', markersize=15)
plt.grid(True)
plt.show()
print "..#### Hey there!! Welcome to the class predictor bot ####.."
error = ""
def bigBang(k, folderName, optimizationParams):
'''
Creates Environment as per the user's preferences
'''
env = Environment(k, folderName, optimizationParams)
env.process()
if len(sys.argv)>1:
'''
If command line arguments are provided, menu is not displayed and processing tarts right away
'''
bigBang(sys.argv[1], sys.argv[2], {"distanceAlgo":"1","dataDistribution":"1"})
else:
'''
If command line arguments are not provided, menu is displayed to accept input from the user
'''
while 1:
'''
Display Menu
'''
optimizationParams = {}
if error == "":
k = raw_input("\nEnter the value of k (Enter 0 if you want me to evaluate how accuracy depends on k) \n")
if int(k) == 0:
maxK = raw_input("\nEnter maximum value of k (Note - Computing time will be higher for high values of k) \n")
optimizationParams["maxK"]=int(maxK)
distanceAlgo = raw_input("\nSelect one of the distance algorithms below:\n1.Intelligent(I will decide distance algorithm for each attribute individually)\n2.Euclidian\n3.Manhattan\n")
optimizationParams["distanceAlgo"]=int(distanceAlgo)
dataDistribution = raw_input("\nDo you want attribute data to be in the same range interval?\n1.Yes\n2.No\n")
optimizationParams["dataDistribution"]=int(dataDistribution)
dirs = next(os.walk('.'))[1]
getFolderName = "\nSelect the folder with data\n"
i=0
folderNames=[]
for fileName in dirs:
i+=1
getFolderName+=repr(i)+"."+fileName+"\n"
folderNames.append(fileName)
folderName = raw_input(getFolderName)
folderName = folderNames[i-1]
try:
'''
Start Processing after accepting user input
'''
bigBang(k, folderName, optimizationParams)
except IOError:
print "#### Data files not correctly formatted or not found!!"
error = "IOError"
exitInput = raw_input("\nEnter e to exit.. Enter any other key to rerun..\n")
if exitInput == 'e':
print "THANK YOU !!"
exit()