-
Notifications
You must be signed in to change notification settings - Fork 6
/
regression.py
161 lines (138 loc) · 6.31 KB
/
regression.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
import pandas as pd
import seaborn
import math
import numpy as np
from sklearn.svm import SVR
from sklearn.ensemble import GradientBoostingRegressor
from sklearn.preprocessing import StandardScaler
from sklearn.metrics import mean_squared_error
from sklearn.externals import joblib
seaborn.set()
def Time_to_label(df_test, y_pred):
top1 = np.zeros(y_pred.shape)
top3 = np.zeros(y_pred.shape)
top50 = np.zeros(y_pred.shape)
current_race_ID = df_test.loc[0].race_id
race_time = []
indices = []
for index, row in df_test.iterrows():
if current_race_ID == row.race_id:
race_time.append(y_pred[index])
indices.append(index)
else:
index_array = [x for _, x in sorted(zip(race_time, indices))]
top1[index_array[0]] = 1
top3[index_array[0]] = 1
top3[index_array[1]] = 1
top3[index_array[2]] = 1
size = len(index_array)
count = 1
for c in index_array:
if count / size <= 0.5:
top50[c] = 1
else:
break
count += 1
indices = [index]
race_time = [y_pred[index]]
current_race_ID = row.race_id
return top1, top3, top50
def write_csv(a, b, c, race_id, horse_id, path, name):
columns = ['RaceID','HorseID','HorseWin','HorseRankTop3','HorseRankTop50Percent']
dataframe = pd.DataFrame({'RaceID':race_id, 'HorseID':horse_id, 'HorseWin':a, 'HorseRankTop3':b, 'HorseRankTop50Percent':c})
dataframe.to_csv("{0}/{1}_predictions.csv".format(path, name), index=False, sep=',', columns=columns)
def Top_1_3_avg(df_test, y_pred):
current_race_ID = df_test.loc[0].race_id
race_time = []
indices = []
num_races = 1
top1 = 0
top3 = 0
avg_rank = 0
for index, row in df_test.iterrows():
if current_race_ID == row.race_id:
race_time.append(y_pred[index])
indices.append(index)
else:
num_races += 1
index_array = [x for _, x in sorted(zip(race_time, indices))]
top_pos = int(df_test.loc[index_array[0]].finishing_position)
avg_rank += top_pos
if top_pos == 1:
top1 += 1
if top_pos <= 3:
top3 += 1
indices = [index]
race_time = [y_pred[index]]
current_race_ID = row.race_id
return float(top1) / num_races, float(top3) / num_races, float(avg_rank) / num_races
print("Loading data...")
df_train = pd.read_csv('data/training.csv')
df_test = pd.read_csv('data/testing.csv')
features = ['actual_weight', 'declared_horse_weight','draw',
'win_odds', 'jockey_ave_rank', 'trainer_ave_rank',
'recent_ave_rank', 'race_distance']
X_train = np.array(df_train[features])
X_test = np.array(df_test[features])
print("Processing timestamps...")
finish_time = df_train['finish_time']
y_train = []
for t in finish_time:
t_arr = t.split('.')
y_train.append(float(t_arr[0])*60 + float(t_arr[1] + '.' + t_arr[2] ))
y_train = np.array(y_train)
finish_time = df_test['finish_time']
y_test = []
for t in finish_time:
t_arr = t.split('.')
y_test.append(float(t_arr[0])*60 + float(t_arr[1] + '.' + t_arr[2] ))
y_test = np.array(y_test)
std_scalar = StandardScaler()
std_scalar.fit(X_train)
X_train_std = std_scalar.transform(X_train)
X_test_std = std_scalar.transform(X_test)
joblib.dump(std_scalar, 'models/std_scalar.pkl')
std_scalar_y = StandardScaler()
std_scalar_y.fit(np.reshape(y_train, (-1, 1)))
y_train_std = std_scalar_y.transform(np.reshape(y_train, (-1, 1))).ravel()
joblib.dump(std_scalar_y, 'models/std_scalar_y.pkl')
#model = SVR(kernel='linear', C = 0.1, epsilon=1) #1.62 without normal
#model = SVR(kernel='linear', C = 10) #1.59 with full data - normalized
#model = SVR(kernel='rbf', C = 5000, gamma= 0.0000001) #1.561 with all data
#model = SVR(kernel='rbf', C = 1000000, gamma= 0.000001) #1.592 with all data - normalized
#model = SVR(kernel='sigmoid', C = 1, gamma=0.01) #1.61 with full data - normalized
#model = SVR(kernel = 'poly', degree=2, C=0.00001) #1.8
#model = SVR(kernel = 'poly', degree=3, coef0=1) #1.565 full data - normalized
#Reason : Fastest to train, same performance as others
#svr_model = SVR(kernel='rbf', C=5000, epsilon=0.1, gamma=0.0000001) 1.59 .227 .496 4.619
print ("model_name RMSE Top_1 Top_3 Average_Rank")
svr_model = SVR(kernel='rbf', C = 5000, epsilon=0.01, gamma= 0.000001)
svr_model.fit(X_train, y_train)
joblib.dump(svr_model, 'models/svr_model.pkl')
svr_pred = svr_model.predict(X_test)
svr_RMSE = math.sqrt(mean_squared_error(y_test, svr_pred))
svr_top1, svr_top3, svr_avg = Top_1_3_avg(df_test, svr_pred)
print ("svr_model, %.3f %.3f %.3f %.3f" %(svr_RMSE, svr_top1, svr_top3, svr_avg))
s_svr_model = SVR(kernel='rbf', C = 1000, epsilon=0.01, gamma= 0.0001)
s_svr_model.fit(X_train_std, y_train_std)
joblib.dump(s_svr_model, 'models/s_svr_model.pkl')
s_svr_pred = s_svr_model.predict(X_test_std)
s_svr_pred = std_scalar_y.inverse_transform(s_svr_pred)
s_svr_RMSE = math.sqrt(mean_squared_error(y_test, s_svr_pred))
s_svr_top1, s_svr_top3, s_svr_avg = Top_1_3_avg(df_test, s_svr_pred)
print ("Scaled svr_model, %.3f %.3f %.3f %.3f" %(s_svr_RMSE, s_svr_top1, s_svr_top3, s_svr_avg))
gbrt_model = GradientBoostingRegressor(loss='ls', learning_rate=0.05, n_estimators=120, max_depth=5, random_state=42)
gbrt_model.fit(X_train, y_train)
joblib.dump(gbrt_model, 'models/gbrt_model.pkl')
gbrt_pred = gbrt_model.predict(X_test)
gbrt_RMSE = math.sqrt(mean_squared_error(y_test, gbrt_pred))
gbrt_top1, gbrt_top3, gbrt_avg = Top_1_3_avg(df_test, gbrt_pred)
print ("gbrt_model, %.3f %.3f %.3f %.3f" %(gbrt_RMSE, gbrt_top1, gbrt_top3, gbrt_avg))
s_gbrt_model = GradientBoostingRegressor(loss='ls', learning_rate=0.05, n_estimators=120, max_depth=5, random_state=42)
s_gbrt_model.fit(X_train_std, y_train_std)
joblib.dump(s_gbrt_model, 'models/s_gbrt_model.pkl')
s_gbrt_pred = s_gbrt_model.predict(X_test_std)
s_gbrt_pred = std_scalar_y.inverse_transform(s_gbrt_pred)
s_gbrt_RMSE = math.sqrt(mean_squared_error(y_test, s_gbrt_pred))
s_gbrt_top1, s_gbrt_top3, s_gbrt_avg = Top_1_3_avg(df_test, s_gbrt_pred)
print ("Scaled gbrt_model, %.3f %.3f %.3f %.3f" %(s_gbrt_RMSE, s_gbrt_top1, s_gbrt_top3, s_gbrt_avg))