/
titanic_03.py
183 lines (137 loc) · 6.04 KB
/
titanic_03.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
# To support both python 2 and python 3
from __future__ import division, print_function, unicode_literals
# Common imports
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
# sklearn model validation and preprocessing imports
from sklearn.base import BaseEstimator, TransformerMixin
from sklearn.preprocessing import Imputer, StandardScaler
from sklearn.pipeline import Pipeline, FeatureUnion
from future_encoders import OrdinalEncoder, OneHotEncoder
from sklearn.metrics import accuracy_score, confusion_matrix
from sklearn.model_selection import cross_validate, GridSearchCV, train_test_split, cross_val_score
# import models to try
from sklearn.linear_model import LogisticRegression
from sklearn.svm import SVC
from sklearn.neighbors import KNeighborsClassifier
from sklearn.tree import DecisionTreeClassifier
from sklearn.ensemble import RandomForestClassifier, GradientBoostingClassifier
# Classificator's names
clfs_name = ["LogisticRegression", "SVC", "KNeighbors", "DecisionTree",
"RandomForest", "GradientBoosting" ]
# Data import
titanic = pd.read_csv('train.csv')
# Drop useless columns and NaN values in the Embarked column
titanic.drop(columns = ["Ticket","Name","Cabin"], inplace=True)
titanic = titanic[titanic["Embarked"].isna()==False]
# Create a class to select numerical or categorical columns
class DataFrameSelector(BaseEstimator, TransformerMixin):
def __init__(self, attribute_names):
self.attribute_names = attribute_names
def fit(self, X, y=None):
return self
def transform(self, X):
return X[self.attribute_names].values
# Numerical features columns names
num_features = ['Pclass', 'Age', 'SibSp', 'Parch', 'Fare']
# Categorical features names
cat_features = ['Sex', 'Embarked']
# Pipeline for numerical features
num_pipeline = Pipeline([
('selector', DataFrameSelector(num_features)),
('imputer', Imputer(strategy="median")),
('std_scaler', StandardScaler()),
])
# Pipeline for "Sex" features
sex_pipeline = Pipeline([
('selector', DataFrameSelector(["Sex"])),
('sex_encoder', OrdinalEncoder()), #OneHotEncoder(sparse=False)
])
# Pipeline for "Embarked" features
emb_pipeline = Pipeline([
('selector', DataFrameSelector(["Embarked"])),
('emb_encoder', OrdinalEncoder()), #OneHotEncoder(sparse=False)
])
# Data preparation pipeline
data_preparation_pipeline = FeatureUnion(transformer_list=[
("num_pipeline", num_pipeline),
("sex_pipeline", sex_pipeline),
("emb_pipeline", emb_pipeline),
], n_jobs=1)
# Prepared data and labels for the fitting
X = data_preparation_pipeline.fit_transform(titanic)
y = titanic["Survived"].values
# Split data in a train and a validation set
X_train, X_val, y_train, y_val = train_test_split(X, y, test_size = .2, random_state=42)
# Initialization of the classificators I whish to try
clfs = []
clfs.append(LogisticRegression())
clfs.append(SVC())
clfs.append(KNeighborsClassifier(n_neighbors=3))
clfs.append(DecisionTreeClassifier())
clfs.append(RandomForestClassifier())
clfs.append(GradientBoostingClassifier())
mean_clfs = []
std_clfs = []
validation_score = []
# Cicle on the classifiers. For each classifier we look for cross validation accuracy score.
# We save the accuracy on the validation set as well
for name, classifier in zip(clfs_name, clfs):
scores = cross_val_score(classifier, X_train, y_train, cv = 7, scoring="accuracy")
print('---------------------------------')
print(name, ':')
print('---------------------------------')
mean_clfs.append(scores.mean())
std_clfs.append(scores.std())
classifier.fit(X_train, y_train)
y_pred = classifier.predict(X_val)
validation_score.append(accuracy_score(y_val,y_pred))
print('Mean: ', scores.mean())
print('Std: ', scores.std())
print('Validation set result: ',validation_score[-1])
# Convert lists into arrays
mean_clfs_num = np.asarray(mean_clfs)
std_clfs_num = np.asarray(std_clfs)
validation_score_num = np.asarray(validation_score)
# Visualization of the model scores
fig, ax1 = plt.subplots(figsize=(10, 6))
fig.canvas.set_window_title('Results for each model')
plt.subplots_adjust(left=0.075, right=0.95, top=0.9, bottom=0.25)
plt.errorbar(np.arange(1,mean_clfs_num.size+1),mean_clfs_num,yerr=2*std_clfs_num,fmt='o',
elinewidth=1, capsize=3, capthick=1)
plt.plot(np.arange(1,mean_clfs_num.size+1),validation_score_num,'rx')
plt.ylim((0,1))
plt.xticks(np.arange(1,mean_clfs_num.size+1), clfs_name, rotation=45)
ax1.set_xlabel('Classificators', labelpad=15, fontsize=15)
ax1.set_ylabel('Mean score', labelpad=15, fontsize=15)
plt.show()
print("The classificator which performs better in average is: ", clfs_name[mean_clfs_num.argmax()])
# Let us try to increment the performance of the GradientBoostingClassifier by tuning its hyperparameters
clf = GradientBoostingClassifier()
param_grid = [
{'loss' : ['deviance', 'exponential'], 'n_estimators': list(range(90,110)),'max_depth': list(range(2,15))},
]
grid_search = GridSearchCV(clf, param_grid, cv=7, scoring='accuracy')
grid_search.fit(X_train, y_train)
print("Results: ")
print(grid_search.best_params_)
clf_best = grid_search.best_estimator_
clf_best.fit(X_train, y_train)
y_train_pred = clf_best.predict(X_train)
accuracy_on_train = accuracy_score(y_train, y_train_pred)
y_val_pred = clf_best.predict(X_val)
accuracy_on_val = accuracy_score(y_val, y_val_pred)
print('Accuracy score on the train set: ', accuracy_on_train)
print('Accuracy score on the validation set: ', accuracy_on_val)
# Output the solution on some .CSV file
give_solution = True
if give_solution:
titanic_test = pd.read_csv('test.csv')
X_test = data_preparation_pipeline.transform(titanic_test)
y_predicted = grid_search.best_estimator_.predict(X_test).reshape(-1,1)
pass_id = titanic_test["PassengerId"].values.reshape(-1,1)
out = np.c_[pass_id, y_predicted]
output = pd.DataFrame(out, columns = ["PassengerId", "Survived"])
output.to_csv(path_or_buf="./Solution.csv", index=False)
print("Output scritto")