时间，把字符串改成日期是原样，数字则从1970开始

本文链接：https://blog.youkuaiyun.com/sphinxrascal168/article/details/104753687

df1['date1'] = df1['DATE'].astype(str)#转换成字符串，挺重要
print(df1.tail(1))
df1['date1'] =pd.to_datetime(df1['date1'])
df1['week1']=df1['date1'].dt.isocalendar().week
df1['year1']=df1['date1'].dt.isocalendar().year#要是不用isocalendar，则是哪一年就哪一年，第一个月的1日容日出错

用了是这样，不然会是2021，week53，因为本质上不是一个函数

TSCODE DATE ROWNUM CLOSE date1 week1 year1
0 SZ000001 20210101 10 14.0 2021-01-01 53 2020
1 SZ000001 20210102 9 11.0 2021-01-02 53 2020
2 SZ000001 20210103 8 13.0 2021-01-03 53 2020

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from sklearn.metrics import mean_squared_error
import math
from math import sqrt
from datetime import datetime
import copy

np.random.seed(2019)
seed = 2019

def rmse(y_pred,y_true):
return sqrt(mean_squared_error(y_pred,y_true))

sample = pd.read_csv('sample_500k_0719.csv')

# drop purcharse_ar column
data = sample.drop('purchase_ar', axis=1)

data.drop('Unnamed: 0', axis=1, inplace=True)

# find time-series characteristics
from scipy import optimize

data.sort_values(by=['becif', 'dt'], inplace=True)

def calc_shifted_ewm(series, alpha, adjust=True):
return series.shift().ewm(alpha=alpha, adjust=adjust).mean()

def find_best_signal(series, adjust=False, eps=10e-5):

def f(alpha):
shifted_ewm = calc_shifted_ewm(series=series, alpha=min(max(alpha, 0), 1), adjust=adjust)
corr = np.mean(np.power(series - shifted_ewm, 2))
return corr

res = optimize.differential_evolution(func=f, bounds=[(0 + eps, 1 - eps)])

return calc_shifted_ewm(series=series, alpha=res['x'][0], adjust=adjust)

roll = data.groupby(['becif', 'dt']).apply(lambda g: calc_shifted_ewm(g['purchase_ar_next'], 0.96))
data['optimized_ewm_by_becif'] = roll.sort_index(level=['becif', 'dt']).values

# get dummies
data =pd.get_dummies(data,columns=['month'])

# train/test
data['par_next_log1p'] = np.log1p(data['purchase_ar_next'])

train = data[data['dt'] < 201900]
train = train.dropna()

test = data[data['dt'] > 201900]

X_train = train.drop('par_next_log1p', axis='columns')
X_test = test.drop('par_next_log1p', axis='columns')
y_train = train['par_next_log1p']

# check datasets
assert X_train.isnull().sum().sum() == 0
assert y_train.isnull().sum() == 0
assert len(X_train) == len(y_train)
assert X_test.isnull().sum().sum() == 0

import lightgbm as lgb
from sklearn import metrics
from sklearn import model_selection

model = lgb.LGBMRegressor(objective='regression',
max_depth=6,
num_leaves=31,
learning_rate=0.01,
n_estimators=1000,
min_child_samples=80,
subsample=0.8,
random_state=np.random.randint(10e6))

n_splits = 6
cv = model_selection.KFold(n_splits=n_splits, shuffle=True, random_state=42)

val_scores = [0] * n_splits

sub = X_test[['becif', 'dt']]
sub['par_pred'] = 0

feature_importances = pd.DataFrame(index=X_train.columns)

for i, (fit_idx, val_idx) in enumerate(cv.split(X_train, y_train)):
X_fit = X_train.iloc[fit_idx]
y_fit = y_train.iloc[fit_idx]
X_val = X_train.iloc[val_idx]
y_val = y_train.iloc[val_idx]

model.fit(X_fit,y_fit,
eval_set=[(X_fit, y_fit), (X_val, y_val)],
eval_names=('fit', 'val'),
eval_metric='l2',
early_stopping_rounds=200,
feature_name=X_fit.columns.tolist(),
verbose=False
)

val_scores[i] = np.sqrt(model.best_score_['val']['l2'])
sub['par_pred'] += model.predict(X_test, num_iteration=model.best_iteration_)
feature_importances[i] = model.feature_importances_

print('Fold {} RMSLE: {:.5f}'.format(i+1, val_scores[i]))