-
Notifications
You must be signed in to change notification settings - Fork 5
/
TitanicRandomForest.py
194 lines (170 loc) · 5.88 KB
/
TitanicRandomForest.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
from __future__ import division
import scipy
from sklearn.ensemble import RandomForestClassifier
import numpy as np
import csv
import sys
# This program will use Random Forest as a machine learning algorithm that predicts
# survival on the Titanic based on a number of characterics about every passenger.
Xtrain=[] # Will contain the characteristic data for each passenger. Xtrain will
# contain info about the passenger's class, sex, age, siblings, parch,
# fare, and cabin, and embark status
Ytrain=[] # WIll contain a binary label for whether each passenger survived
Xtest=[] # Will contain the characteristic data for the passengers in the test set
numTrainExamples = 891
numTestExamples = 418
# The next few functions are going to serve as preprocessing steps for the different
# features found in the input. Preprocessing will mainly end up being mean normalization
# and filling in blank values. The main process function (below) just applies mean
# normalization to the values in the set. However, we need specific process functions
# for characteristics like gender, where we have to assign a number value to the
# characteristic of male or female.
def process(pclass):
pclass = [float(x) for x in pclass]
mean = sum(pclass) / float(len(pclass))
my_range = max(pclass) - min(pclass)
pclass = [(x-mean)/my_range for x in pclass]
return pclass
def processGender(gender):
genderToNum=[]
for x in gender:
if (x == "male"):
genderToNum.append(-1)
else: # x is female
genderToNum.append(1)
return genderToNum
def processAgeorFare(age):
ageWithoutBlanks=[]
for x in age:
if (x != ""):
ageWithoutBlanks.append(x)
ageWithoutBlanks = [float(x) for x in ageWithoutBlanks]
mean = sum(ageWithoutBlanks) / float(len(ageWithoutBlanks))
my_range = max(ageWithoutBlanks) - min(ageWithoutBlanks)
#median = np.median(ageWithoutBlanks)
for x in range(0,len(age)):
if (age[x] == ""): # If there is a blank value, then just set it to mean
age[x] = mean
age = [float(x) for x in age]
age = [(x-mean)/my_range for x in age]
return age
def processCabin(cabin):
cabinToNum=[]
for x in gender:
if (x == ""):
cabinToNum.append(-1)
else:
cabinToNum.append(1)
return cabinToNum
def processName(name):
nameToNum=[]
rareTitles=['Dona', 'Lady', 'the Countess','Capt', 'Col', 'Don',
'Dr', 'Major', 'Rev', 'Sir', 'Jonkheer']
skip=False
for x in name:
for y in rareTitles:
if ((y in x) and (skip==False)):
nameToNum.append(1)
skip=True
if (skip==True):
skip=False
else:
nameToNum.append(-1)
return nameToNum
def processEmbarked(embarked):
embarkedToNum=[]
for x in embarked:
if (x == "S"):
embarkedToNum.append(-1)
elif (x == "Q"):
embarkedToNum.append(1)
else: # x is C or blank
embarkedToNum.append(0)
return embarkedToNum
# First job is to read in the data from the training data that Kaggle provides. This
# training data is in the form of a csv file. This CSV file should be in the same
# directory as this program.
# We want to skip the first row in the csv file because it just
# contains column header
skip = True
train_file = open('train.csv')
csv_file = csv.reader(train_file)
# Creating temporary lists where we store data for each feature/characteristic
name,gender,Pclass,age,sibSP,parch,fare,cabin,embarked = ([] for i in range(9))
for row in csv_file:
if (skip == True):
skip = False
continue
# Filling lists with values from train.csv
Ytrain.append(row[1])
Pclass.append(row[2])
name.append(row[3])
gender.append(row[4])
age.append(row[5])
sibSP.append(row[6])
parch.append(row[7])
fare.append(row[9])
cabin.append(row[10])
embarked.append(row[11])
# Processing each feature list
Pclass = process(Pclass)
name = processName(name)
gender = processGender(gender)
age = processAgeorFare(age)
sibSP = process(sibSP)
parch = process(parch)
fare = processAgeorFare(fare)
cabin = processCabin(cabin)
embarked = processEmbarked(embarked)
# Adding values from previous feature lists to one large Xtrain list of lists
for x in range(0,numTrainExamples):
Xtrain.append([Pclass[x],name[x],gender[x],age[x],sibSP[x],
parch[x],fare[x],cabin[x],embarked[x]])
# Repeating same process for test file (except we don't know Ytest)
skip = True
test_file = open('test.csv')
csv_file2 = csv.reader(test_file)
name,gender,Pclass,age,sibSP,parch,fare,cabin,embarked = ([] for i in range(9))
for row in csv_file2:
if (skip == True):
skip = False
continue
Pclass.append(row[1])
name.append(row[2])
gender.append(row[3])
age.append(row[4])
sibSP.append(row[5])
parch.append(row[6])
fare.append(row[8])
cabin.append(row[9])
embarked.append(row[10])
Pclass = process(Pclass)
name = processName(name)
gender = processGender(gender)
age = processAgeorFare(age)
sibSP = process(sibSP)
parch = process(parch)
fare = processAgeorFare(fare)
cabin = processCabin(cabin)
embarked = processEmbarked(embarked)
for x in range(0,numTestExamples):
Xtest.append([Pclass[x],name[x],gender[x],age[x],sibSP[x],
parch[x],fare[x],cabin[x],embarked[x]])
# Representing the list of lists as numpy arrays so that we can
# use the different scikit functions.
Xtrain = np.asarray(Xtrain)
Xtest = np.asarray(Xtest)
Ytrain = np.asarray(Ytrain)
print 'Fitting Random Forest'
clf = RandomForestClassifier(n_estimators=100)
clf.fit(Xtrain, Ytrain)
results = np.ones((numTestExamples,2))
counter = numTrainExamples + 1
print 'Predicting outputs for testing dataset'
for x in range(0,numTestExamples):
results[x,1] = (clf.predict(Xtest[x]))[0]
results[x,0] = counter
counter = counter + 1
#Saving predictions into a test file that can be uploaded to Kaggle
#NOTE: You have to add a header row before submitting the txt file
np.savetxt('result.csv', results, delimiter=',', fmt = '%i')