# -*- coding: utf-8 -*-

"""

Author: Ying

Date: 2018/09

"""

import numpy as np

import pandas as pd

import xgboost as xgb

import constant

from tools import build_datasets, build_baseline_mae

# Step 1: 读取数据集

X_train, X_test, y_train, y_test = build_datasets()

# Step 2: 构建DMatrix - dtain & dtest

dtrain = xgb.DMatrix(X_train, label=y_train)

dtest = xgb.DMatrix(X_test, label=y_test)

# Step 3: 构建baseline model

build_baseline_mae(y_train, y_test)

print('================\n')

# Step 4: Tuning an XGBoost model

# The params dictionary

params = {

}

num_boost_round=999

# - 4.1 simple train with linear model

# model = xgb.train(

# params,

# dtrain,

# num_boost_round=num_boost_round,

# evals=[(dtest, "Test")],

# early_stopping_rounds=10

# )

#

# print("Best MAE: {:.5f} with {} rounds".format(

# model.best_score,

# model.best_iteration+1))

# print('================\n')

# - 4.2 xgb cross-validation

# cv_results = xgb.cv(

# params,

# dtrain,

# num_boost_round=num_boost_round,

# seed=42,

# nfold=5,

# metrics={'mae'},

# early_stopping_rounds=10,

# verbose_eval=True,

# )

#

# print(cv_results)

# print('================\n')

# - 4.3 Parameters max_depth and min_child_weight

# - max_depth: 9

# - min_child_weight: 3

#

# gridsearch_params = [

# (max_depth, min_child_weight)

# for max_depth in range(1, 4)

# for min_child_weight in range(1, 3)

# ]

#

# params['silent'] = 1

# # Define initial best params and MAE

# min_mae = float("Inf")

# best_params = None

# for max_depth, min_child_weight in gridsearch_params:

# print("CV with max_depth={}, min_child_weight={}".format(

# max_depth,

# min_child_weight))

#

# # Update our parameters

# params['max_depth'] = max_depth

# params['min_child_weight'] = min_child_weight

#

# # Run CV

# cv_results = xgb.cv(

# params,

# dtrain,

# num_boost_round=num_boost_round,

# seed=42,

# nfold=5,

# metrics={'mae'},

# early_stopping_rounds=10

# )

#

#

# # Update best MAE

# mean_mae = cv_results['test-mae-mean'].min()

# boost_rounds = cv_results['test-mae-mean'].idxmin()

# print("\tMAE {} for {} rounds".format(mean_mae, boost_rounds))

# if mean_mae < min_mae:

# min_mae = mean_mae

# best_params = (max_depth,min_child_weight)

#

# print("Best params: {}, {}, MAE: {}".format(best_params[0], best_params[1], min_mae))

# print('================\n')

# - 4.4 Parameters subsample and colsample_bytree

# - subsample: 1

# - colsample_bytree: 0.8

#

# gridsearch_params = [

# (subsample, colsample)

# for subsample in [i/10. for i in range(5,11)]

# for colsample in [i/10. for i in range(5,11)]

# ]

#

# min_mae = float("Inf")

# best_params = None

# params['silent'] = 1

#

# # We start by the largest values and go down to the smallest

# for subsample, colsample in reversed(gridsearch_params):

# print("CV with subsample={}, colsample={}".format(

# subsample,

# colsample))

#

# # We update our parameters

# params['subsample'] = subsample

# params['colsample_bytree'] = colsample

#

# # Run CV

# cv_results = xgb.cv(

# params,

# dtrain,

# num_boost_round=num_boost_round,

# seed=42,

# nfold=5,

# metrics={'mae'},

# early_stopping_rounds=10

# )

#

# # Update best score

# mean_mae = cv_results['test-mae-mean'].min()

# boost_rounds = cv_results['test-mae-mean'].argmin()

# print("\tMAE {} for {} rounds".format(mean_mae, boost_rounds))

# if mean_mae < min_mae:

# min_mae = mean_mae

# best_params = (subsample,colsample)

#

# print("Best params: {}, {}, MAE: {}".format(best_params[0], best_params[1], min_mae))

# print('================\n')

# - 4.5 Parameter ETA

# - ETA: 0.1 (0.034535 with 281 rounds)

# - ETA: 0.05 (0.034252 with 592 rounds)

# This can take some time…

min_mae = float("Inf")

best_params = None

params['silent'] = 1

for eta in [.3, .2, .1, .05, .01, .005]:

print("CV with eta={}".format(eta))

# We update our parameters

params['eta'] = eta

# Run and time CV

cv_results = xgb.cv(

params,

dtrain,

num_boost_round=num_boost_round,

seed=42,

nfold=5,

metrics=['mae'],

early_stopping_rounds=10

)

# Update best score

mean_mae = cv_results['test-mae-mean'].min()

boost_rounds = cv_results['test-mae-mean'].argmin()

print("\tMAE {} for {} rounds\n".format(mean_mae, boost_rounds))

if mean_mae < min_mae:

min_mae = mean_mae

best_params = eta

print("Best params: {}, MAE: {}".format(best_params, min_mae))

print('================\n')