-
Notifications
You must be signed in to change notification settings - Fork 0
/
bikes_challenge_kaggle.py
283 lines (202 loc) · 7.2 KB
/
bikes_challenge_kaggle.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
# -*- coding: utf-8 -*-
"""
How Many Bikes ?
Kaggle Challenge user name: alextuil
Alexis Tuil
12/16/2016
"""
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from patsy import dmatrices
from sklearn.model_selection import KFold
from sklearn.linear_model import LinearRegression, Ridge
from sklearn import svm
from sklearn.preprocessing import scale, StandardScaler
from sklearn.metrics import make_scorer
def RMSLE(y, y_pred):
y_pred[y_pred < 0] = 0
return np.sqrt( np.mean((np.log(y_pred + 1) - np.log(y + 1))**2) )
dataset = pd.read_csv('train.csv', header=0)
y, X = dmatrices('cnt ~ 0 + yr + season + mnth + hr + holiday + weekday + workingday + weathersit + temp + atemp + hum + windspeed',dataset, return_type='dataframe')
# set up a k-fold cross validation on the training set
kf = KFold(n_splits=5)
# linear regression on the full training data
lr = LinearRegression()
lr.fit(X,y)
y_pred = lr.predict(X)
print(RMSLE(y,y_pred))
# 1.300951
# linear regression cross-validated
scores = []
y = np.ravel(y)
Xm = X.as_matrix()
for train, test in kf.split(Xm):
model = LinearRegression()
model.fit(Xm[train],y[train])
y_pred = model.predict(Xm[test])
scores.append(RMSLE(y[test],y_pred))
print(np.mean(scores))
# 1.36903506661
# ridge regression on the full training data
a_range = np.arange(0,3000,10)
a_score = []
for a in a_range:
rdg = Ridge(alpha=a)
rdg.fit(X,y)
y_pred = rdg.predict(X)
a_score.append(RMSLE(y,y_pred))
plt.plot(a_range, a_score)
plt.xlabel('Ridge Parameter')
plt.ylabel('RMSLE score')
plt.title('Ridge regression on the full training set')
# best for alpha=750
rdg = Ridge(alpha=750)
rdg.fit(X,y)
print(RMSLE(y,rdg.predict(X)))
# RMSLE = 1.23244788655
# c) ridge regression cross-validated
a_range = np.arange(0,2000,10)
a_score = []
for a in a_range:
scores = []
for train, test in kf.split(Xm):
rdg = Ridge(alpha=a)
rdg.fit(Xm[train],y[train])
y_pred = rdg.predict(Xm[test])
scores.append(RMSLE(y[test],y_pred))
a_score.append(np.mean(scores))
plt.plot(a_range, a_score)
plt.xlabel('Ridge Parameter')
plt.ylabel('RMSLE score')
plt.title('Ridge regression cross-validated')
# best for alpha=500
scores = []
for train, test in kf.split(Xm):
rdg = Ridge(alpha=500)
rdg.fit(Xm[train],y[train])
scores.append(RMSLE(y[test],rdg.predict(Xm[test])))
print(np.mean(scores))
# RMSLE = 1.24320579537
# ridge regression on the full training data with scaling
X_scaled = scale(X)
# when training then evaluating on the full training set
a_range = np.arange(0,6000,10)
a_score = []
for a in a_range:
rdg = Ridge(alpha=a)
rdg.fit(X_scaled,y)
y_pred = rdg.predict(X_scaled)
a_score.append(RMSLE(y,y_pred))
plt.plot(a_range, a_score)
plt.xlabel('Ridge Parameter')
plt.ylabel('RMSLE score')
plt.title('Ridge regression on the full training set and features scaled')
# best for alpha=3500
rdg = Ridge(alpha=3500)
rdg.fit(X_scaled,y)
print(RMSLE(y,rdg.predict(X_scaled)))
# RMSLE = 1.26469309761
# ridge regression cross-validated with scaling
a_range = np.arange(0,6000,1)
a_score = []
for a in a_range:
scores = []
for train, test in kf.split(Xm):
rdg = Ridge(alpha=a)
scaler = StandardScaler().fit(Xm[train])
Xm_train_scaled = scaler.transform(Xm[train])
rdg.fit(Xm_train_scaled,y[train])
Xm_test_scaled = scaler.transform(Xm[test])
y_pred = rdg.predict(Xm_test_scaled)
scores.append(RMSLE(y[test],y_pred))
a_score.append(np.mean(scores))
plt.plot(a_range, a_score)
plt.xlabel('Ridge Parameter')
plt.ylabel('RMSLE score')
plt.title('Ridge regression cross-validated and features scaled')
# best for alpha = 3500
scores = []
for train, test in kf.split(Xm):
rdg = Ridge(alpha=3500)
scaler = StandardScaler().fit(Xm[train])
Xm_train_scaled = scaler.transform(Xm[train])
rdg.fit(Xm_train_scaled,y[train])
Xm_test_scaled = scaler.transform(Xm[test])
y_pred = rdg.predict(Xm_test_scaled)
scores.append(RMSLE(y[test],y_pred))
print(np.mean(scores))
# RMSLE = 1.27989927838
dataset_test = pd.read_csv('test.csv', header=0)
X_test = dataset_test[['season','yr','mnth','hr','holiday','weekday','workingday','weathersit','temp','atemp','hum','windspeed']]
model = linear_model.Ridge(alpha=500)
model.fit(X,y)
y_pred = model.predict(X_test)
for i in range(len(y_pred)):
y_pred[i] = math.floor(max(0,y_pred[i]))
d = {'id': np.ravel(dataset_test[['instant']]), 'prediction': y_pred}
df = pd.DataFrame(d,)
df.to_csv("submission-1.csv", index=False)
# sent to Kaggle Leaderboard RMLSE = 1.29616
RMSLE_scoring = make_scorer(score_func=RMSLE, greater_is_better=False)
# (a) Linear kernel
c_range = np.logspace(-6, 5, 9)
scores = []
for c in c_range:
rmlse_scores = []
model = svm.LinearSVR(C=c)
score = - cross_val_score(model, X, y, cv=5, scoring=RMSLE_scoring)
scores.append(np.mean(score))
plt.semilogx(c_range, scores)
plt.xlabel('C')
plt.ylabel('RMLSE score')
plt.title('Linear kernel')
plt.show()
# Gaussian RDB kernel
c_range = np.logspace(0, 2, 3)
gamma_range = np.logspace(-3, 0, 4)
scores = pd.DataFrame()
for c in c_range:
for g in gamma_range:
rmlse_scores = []
model = svm.SVR(kernel='rbf', gamma=g, C=c)
score = -cross_val_score(model, X, y, cv=5, scoring=RMSLE_scoring)
res = pd.DataFrame({'C': c, 'gamma': g, 'RMSLE': np.mean(score)}, index=np.arange(1))
scores = scores.append(res)
# Test Data
model = svm.SVR(kernel='rbf', gamma=0.1, C=100)
model.fit(X,y)
y_pred = model.predict(X_test)
for i in range(len(y_pred)):
y_pred[i] = math.floor(max(0,y_pred[i]))
d = {'id': np.ravel(dataset_test[['instant']]), 'prediction': y_pred}
df = pd.DataFrame(d,)
df.to_csv("submission-2.csv", index=False)
# sent to Kaggle Leaderboard RMLSE = 0.61700
# Random Forest for different parameters
max_depth_range = [10,15,20,None]
min_samples_leaf_range= [2,5,10,20]
scores = pd.DataFrame()
for md in max_depth_range:
for msl in min_samples_leaf_range:
rmlse_scores = []
model = RandomForestRegressor(n_estimators=500, max_depth=md, min_samples_leaf=msl, max_features='auto', n_jobs=-1, random_state=0)
score = -cross_val_score(model, X, y, cv=5, scoring=RMSLE_scoring)
res = pd.DataFrame({'Max Depth': md, 'Min Sample Leaf': msl, 'RMSLE': np.mean(score)}, index=np.arange(1))
scores = scores.append(res)
print(scores)
# best for max_depth = None and min_sample_leaf = 2
model = RandomForestRegressor(n_estimators=500, max_depth=None, min_samples_leaf=2, max_features='auto', n_jobs=-1, random_state=0)
score = - cross_val_score(model, X, y, cv=5, scoring=RMSLE_scoring)
print(np.mean(score))
# RMLSE = 0.470432421612
# Test Data
model = RandomForestRegressor(n_estimators=500, max_depth=None, min_samples_leaf=2, max_features='auto', n_jobs=-1, random_state=0)
model.fit(X,y)
y_pred = model.predict(X_test)
for i in range(len(y_pred)):
y_pred[i] = math.floor(max(0,y_pred[i]))
d = {'id': np.ravel(dataset_test[['instant']]), 'prediction': y_pred}
df = pd.DataFrame(d,)
df.to_csv("submission-3.csv", index=False)
# sent to Kaggle Leaderboard RMLSE = 0.47329