-
Notifications
You must be signed in to change notification settings - Fork 19
/
example04_housing_validation.py
101 lines (90 loc) · 3.53 KB
/
example04_housing_validation.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
"""
Copyright (c) 2021 Olivier Sprangers as part of Airlab Amsterdam
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
https://github.com/elephaint/pgbm/blob/main/LICENSE
"""
#%% Load packages
import torch
from pgbm.torch import PGBM
from sklearn.model_selection import train_test_split
from sklearn.datasets import fetch_california_housing
import matplotlib.pyplot as plt
#%% Objective for pgbm
def mseloss_objective(yhat, y, sample_weight=None):
gradient = (yhat - y)
hessian = torch.ones_like(yhat)
return gradient, hessian
def rmseloss_metric(yhat, y, sample_weight=None):
loss = (yhat - y).pow(2).mean().sqrt()
return loss
#%% Load data
X, y = fetch_california_housing(return_X_y=True)
#%% Parameters
params = {'min_split_gain':0,
'min_data_in_leaf':2,
'max_leaves':8,
'max_bin':64,
'learning_rate':0.1,
'n_estimators':2000,
'verbose':2,
'early_stopping_rounds':100,
'feature_fraction':1,
'bagging_fraction':1,
'seed':1,
'reg_lambda':1,
'device':'gpu',
'gpu_device_id':0,
'derivatives':'exact',
'distribution':'normal'}
n_forecasts = 1000
n_splits = 2
base_estimators = 2000
#%% Validation loop
rmse, crps = torch.zeros(n_splits), torch.zeros(n_splits)
for i in range(n_splits):
print(f'Fold {i+1}/{n_splits}')
# Split for model validation
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.1, random_state=i)
X_train_val, X_val, y_train_val, y_val = train_test_split(X_train, y_train, test_size=0.2, random_state=i)
# Build datasets
train_data = (X_train, y_train)
train_val_data = (X_train_val, y_train_val)
valid_data = (X_val, y_val)
# Train to retrieve best iteration
print('PGBM Validating on partial dataset...')
params['n_estimators'] = base_estimators
model = PGBM()
model.train(train_val_data, objective=mseloss_objective, metric=rmseloss_metric, valid_set=valid_data, params=params)
# Set iterations to best iteration
params['n_estimators'] = model.best_iteration
# Retrain on full set
print('PGBM Training on full dataset...')
model = PGBM()
model.train(train_data, objective=mseloss_objective, metric=rmseloss_metric, params=params)
#% Predictions
print('PGBM Prediction...')
yhat_point = model.predict(X_test)
yhat_dist = model.predict_dist(X_test, n_forecasts=n_forecasts)
# Scoring
rmse[i] = model.metric(yhat_point.cpu(), y_test)
crps[i] = model.crps_ensemble(yhat_dist.cpu(), y_test).mean()
# Print scores current fold
print(f'RMSE Fold {i+1}, {rmse[i]:.2f}')
print(f'CRPS Fold {i+1}, {crps[i]:.2f}')
# Print final scores
print(f'RMSE {rmse.mean():.2f}+-{rmse.std():.2f}')
print(f'CRPS {crps.mean():.2f}+-{crps.std():.2f}')
#%% Plot all samples
plt.plot(y_test, 'o', label='Actual')
plt.plot(yhat_point.cpu(), 'ko', label='Point prediction PGBM')
plt.plot(yhat_dist.cpu().max(dim=0).values, 'k--', label='Max bound PGBM')
plt.plot(yhat_dist.cpu().min(dim=0).values, 'k--', label='Min bound PGBM')
plt.legend()