import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import dateutil.easter as easter
import holidays
# Extract Holidays
easter_dict = {}
for year in range(2015,2020):
easter_dict[year] = easter.easter(year)
moms_day = {2015: pd.Timestamp(('2015-5-31')),
2016: pd.Timestamp(('2016-5-29')),
2017: pd.Timestamp(('2017-5-28')),
2018: pd.Timestamp(('2018-5-27')),
2019: pd.Timestamp(('2019-5-26'))}
wed_june = {2015: pd.Timestamp(('2015-06-24')),
2016: pd.Timestamp(('2016-06-29')),
2017: pd.Timestamp(('2017-06-28')),
2018: pd.Timestamp(('2018-06-27')),
2019: pd.Timestamp(('2019-06-26'))}
sun_nov = {2015: pd.Timestamp(('2015-11-1')),
2016: pd.Timestamp(('2016-11-6')),
2017: pd.Timestamp(('2017-11-5')),
2018: pd.Timestamp(('2018-11-4')),
2019: pd.Timestamp(('2019-11-3'))}
class Preprocess:
def __init__(self, gdp_path, easter_dict):
self.gdp_path = gdp_path
self.easter_dict = easter_dict
def import_data(self, path):
self.df = pd.read_csv(path, parse_dates=['date'], index_col=0)
self.index = self.df.index
def get_gdp(self, path):
gdp = pd.read_csv(path)
gdp.columns = gdp.columns.str.title().str.replace('Gdp_','')
gdp = gdp.set_index('Year').stack(0).reset_index()
gdp.columns = ['year','country','gdp']
return gdp
def create_date_features(self, df):
df['day_of_year'] = df.date.dt.day_of_year
df['day_of_month'] = df.date.dt.day
df['day_of_week'] = df.date.dt.weekday
# df['month'] = df.date.dt.month
# df['quarter'] = df.date.dt.quarter
df['year'] = df.date.dt.year
# df['period'] = df.date.dt.to_period('M')
return df
def add_country_holidays(self, df):
years = [2015, 2016, 2017, 2018, 2019]
finland = pd.DataFrame([dict(date = date,
finland_holiday = event,
country= 'Finland') for date, event in holidays.Finland(years=years).items()])
norway = pd.DataFrame([dict(date = date,
norway_holiday = event,
country= 'Norway') for date, event in holidays.Norway(years=years).items()])
sweden = pd.DataFrame([dict(date = date,
sweden_holiday = event.replace(", Söndag", ""),
country= 'Sweden') for date, event in holidays.Sweden(years=years).items() if event != 'Söndag'])
finland['date'] = finland['date'].astype("datetime64")
norway['date'] = norway['date'].astype("datetime64")
sweden['date'] = sweden['date'].astype("datetime64")
return (df.merge(finland, on = ['date', 'country'], how = 'left')
.merge(norway, on = ['date', 'country'], how = 'left')
.merge(sweden, on = ['date', 'country'], how = 'left'))
def get_specific_dates_features(self, df, easter_dict):
return (df.assign(easter = lambda x: x.year.map(easter_dict).astype('datetime64'),
moms_day = lambda x: x.year.map(moms_day).astype('datetime64'),
wed_jun = lambda x: x.year.map(wed_june).astype('datetime64'),
sun_nov = lambda x: x.year.map(sun_nov).astype('datetime64'),
days_from_easter = lambda x: (x.date - x.easter).dt.days.clip(-5, 65),
days_from_mom = lambda x: (x.date - x.moms_day).dt.days.clip(-1, 9),
days_from_wed = lambda x: (x.date - x.wed_jun).dt.days.clip(-5, 5),
days_from_sun = lambda x: (x.date - x.sun_nov).dt.days.clip(-1, 9),
)).drop(columns = ['easter','moms_day','wed_jun','sun_nov'])
def join_gdp(self, df):
return df.merge(self.gdp, on = ['country','year'], how = 'left')
def feature_engineering(self, df):
df['log_gdp'] = np.log(df.gdp)
days_cats = [df.day_of_week < 4, df.day_of_week == 4, df.day_of_week > 4]
days = ['week','friday','weekends']
df['week'] = np.select(days_cats, days)
return df
def __call__(self, path, name = 'Train'):
self.import_data(path)
self.gdp = self.get_gdp(self.gdp_path).set_index('year')
out = (self.df.pipe(self.create_date_features)
.pipe(self.join_gdp)
.pipe(self.feature_engineering)
.pipe(self.add_country_holidays)
.pipe(self.get_specific_dates_features, self.easter_dict))
out.index = self.index
print(f'{name} Set created with {out.shape[1]} features')
return out
GDP_PATH = '../inputs/GDP_data_2015_to_2019_Finland_Norway_Sweden.csv'
TRAIN_PATH = '../inputs/train.csv'
TEST_PATH = '../inputs/test.csv'
feature_engineer = Preprocess(GDP_PATH, easter_dict)
df_train = feature_engineer(TRAIN_PATH)
df_test = feature_engineer(TEST_PATH, name = 'Test')
Train Set created with 19 features
Test Set created with 18 features
df_train
| date | country | store | product | num_sold | day_of_year | day_of_month | day_of_week | year | gdp | log_gdp | week | finland_holiday | norway_holiday | sweden_holiday | days_from_easter | days_from_mom | days_from_wed | days_from_sun | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| row_id | |||||||||||||||||||
| 0 | 2015-01-01 | Finland | KaggleMart | Kaggle Mug | 329 | 1 | 1 | 3 | 2015 | 234.440 | 5.457200 | week | Uudenvuodenpäivä | NaN | NaN | -5 | -1 | -5 | -1 |
| 1 | 2015-01-01 | Finland | KaggleMart | Kaggle Hat | 520 | 1 | 1 | 3 | 2015 | 234.440 | 5.457200 | week | Uudenvuodenpäivä | NaN | NaN | -5 | -1 | -5 | -1 |
| 2 | 2015-01-01 | Finland | KaggleMart | Kaggle Sticker | 146 | 1 | 1 | 3 | 2015 | 234.440 | 5.457200 | week | Uudenvuodenpäivä | NaN | NaN | -5 | -1 | -5 | -1 |
| 3 | 2015-01-01 | Finland | KaggleRama | Kaggle Mug | 572 | 1 | 1 | 3 | 2015 | 234.440 | 5.457200 | week | Uudenvuodenpäivä | NaN | NaN | -5 | -1 | -5 | -1 |
| 4 | 2015-01-01 | Finland | KaggleRama | Kaggle Hat | 911 | 1 | 1 | 3 | 2015 | 234.440 | 5.457200 | week | Uudenvuodenpäivä | NaN | NaN | -5 | -1 | -5 | -1 |
| ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... |
| 26293 | 2018-12-31 | Sweden | KaggleMart | Kaggle Hat | 823 | 365 | 31 | 0 | 2018 | 555.455 | 6.319788 | week | NaN | NaN | Nyårsafton | 65 | 9 | 5 | 9 |
| 26294 | 2018-12-31 | Sweden | KaggleMart | Kaggle Sticker | 250 | 365 | 31 | 0 | 2018 | 555.455 | 6.319788 | week | NaN | NaN | Nyårsafton | 65 | 9 | 5 | 9 |
| 26295 | 2018-12-31 | Sweden | KaggleRama | Kaggle Mug | 1004 | 365 | 31 | 0 | 2018 | 555.455 | 6.319788 | week | NaN | NaN | Nyårsafton | 65 | 9 | 5 | 9 |
| 26296 | 2018-12-31 | Sweden | KaggleRama | Kaggle Hat | 1441 | 365 | 31 | 0 | 2018 | 555.455 | 6.319788 | week | NaN | NaN | Nyårsafton | 65 | 9 | 5 | 9 |
| 26297 | 2018-12-31 | Sweden | KaggleRama | Kaggle Sticker | 388 | 365 | 31 | 0 | 2018 | 555.455 | 6.319788 | week | NaN | NaN | Nyårsafton | 65 | 9 | 5 | 9 |
26298 rows × 19 columns
Training Simple Model
X = df_train.drop(columns = ['gdp','date','num_sold'])
y = df_train.num_sold
X_test = df_test.drop(columns = ['gdp','date'])
from feature_engine.creation import CyclicalTransformer
# New class allowing to add a suffix to differentiate different n values.
class CyclicalTransformerV2(CyclicalTransformer):
def __init__(self, suffix = None, **kwargs):
super().__init__(**kwargs)
self.suffix = suffix
def transform(self, X):
X = super().transform(X)
if self.suffix is not None:
transformed_names = X.filter(regex = r'sin$|cos$').columns
new_names = {name: name + self.suffix for name in transformed_names}
X.rename(columns=new_names, inplace=True)
return X
from sklearn.pipeline import Pipeline
from feature_engine.encoding import OneHotEncoder, OrdinalEncoder
from feature_engine.creation import CyclicalTransformer, CombineWithReferenceFeature
from sklearn.preprocessing import StandardScaler
from sklearn.linear_model import Ridge
from feature_engine.selection import DropFeatures
from feature_engine.wrappers import SklearnTransformerWrapper
from feature_engine.imputation import CategoricalImputer
scaler = SklearnTransformerWrapper(StandardScaler())
cyc_1 = CyclicalTransformerV2(variables = ['day_of_year'],
max_values = {'day_of_year':365},
suffix = '_1')
cyc_2 = CyclicalTransformerV2(variables = ['day_of_year'],
max_values = {'day_of_year':365/2,},
suffix = '_2')
linear_prep = Pipeline(steps = [
('cat_imp', CategoricalImputer()),
('ohe', OneHotEncoder(drop_last=True)),
('cyc1', cyc_1),
('cyc2', cyc_2),
('combo', CombineWithReferenceFeature(variables_to_combine=['day_of_year_sin_1', 'day_of_year_cos_1',
'day_of_year_sin_2', 'day_of_year_cos_2'],
reference_variables=['product_Kaggle Mug', 'product_Kaggle Hat'],
operations = ['mul'])),
('drop', DropFeatures(features_to_drop=['year'],#'day_of_year','day_of_week','day_of_month',,]
# 'store_KaggleMart','product_Kaggle Sticker', 'week_week','country_Sweden']
)),
('sc', scaler)
])
log_y = np.log(y)
linear_prep.fit_transform(X)
| day_of_year | day_of_month | day_of_week | log_gdp | days_from_easter | days_from_mom | days_from_wed | days_from_sun | country_Finland | country_Norway | ... | day_of_year_sin_2 | day_of_year_cos_2 | day_of_year_sin_1_mul_product_Kaggle Mug | day_of_year_cos_1_mul_product_Kaggle Mug | day_of_year_sin_2_mul_product_Kaggle Mug | day_of_year_cos_2_mul_product_Kaggle Mug | day_of_year_sin_1_mul_product_Kaggle Hat | day_of_year_cos_1_mul_product_Kaggle Hat | day_of_year_sin_2_mul_product_Kaggle Hat | day_of_year_cos_2_mul_product_Kaggle Hat | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| row_id | |||||||||||||||||||||
| 0 | -1.727310 | -1.673805 | -0.001026 | -1.503796 | -1.503789 | -1.192261 | -1.036474 | -0.421954 | 1.414214 | -0.707107 | ... | 0.048663 | 1.411927 | 0.042169 | 2.447731 | 0.084325 | 2.446645 | -0.000010 | -0.000559 | -0.000019 | -0.000558 |
| 1 | -1.727310 | -1.673805 | -0.001026 | -1.503796 | -1.503789 | -1.192261 | -1.036474 | -0.421954 | 1.414214 | -0.707107 | ... | 0.048663 | 1.411927 | -0.000010 | -0.000559 | -0.000019 | -0.000558 | 0.042169 | 2.447731 | 0.084325 | 2.446645 |
| 2 | -1.727310 | -1.673805 | -0.001026 | -1.503796 | -1.503789 | -1.192261 | -1.036474 | -0.421954 | 1.414214 | -0.707107 | ... | 0.048663 | 1.411927 | -0.000010 | -0.000559 | -0.000019 | -0.000558 | -0.000010 | -0.000559 | -0.000019 | -0.000558 |
| 3 | -1.727310 | -1.673805 | -0.001026 | -1.503796 | -1.503789 | -1.192261 | -1.036474 | -0.421954 | 1.414214 | -0.707107 | ... | 0.048663 | 1.411927 | 0.042169 | 2.447731 | 0.084325 | 2.446645 | -0.000010 | -0.000559 | -0.000019 | -0.000558 |
| 4 | -1.727310 | -1.673805 | -0.001026 | -1.503796 | -1.503789 | -1.192261 | -1.036474 | -0.421954 | 1.414214 | -0.707107 | ... | 0.048663 | 1.411927 | -0.000010 | -0.000559 | -0.000019 | -0.000558 | 0.042169 | 2.447731 | 0.084325 | 2.446645 |
| ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... |
| 26293 | 1.724934 | 1.735250 | -1.500642 | 1.260881 | 0.775294 | 0.854617 | 0.982577 | 2.458770 | -0.707107 | -0.707107 | ... | -0.000033 | 1.412765 | -0.000010 | -0.000559 | -0.000019 | -0.000558 | -0.000010 | 2.448094 | -0.000019 | 2.448096 |
| 26294 | 1.724934 | 1.735250 | -1.500642 | 1.260881 | 0.775294 | 0.854617 | 0.982577 | 2.458770 | -0.707107 | -0.707107 | ... | -0.000033 | 1.412765 | -0.000010 | -0.000559 | -0.000019 | -0.000558 | -0.000010 | -0.000559 | -0.000019 | -0.000558 |
| 26295 | 1.724934 | 1.735250 | -1.500642 | 1.260881 | 0.775294 | 0.854617 | 0.982577 | 2.458770 | -0.707107 | -0.707107 | ... | -0.000033 | 1.412765 | -0.000010 | 2.448094 | -0.000019 | 2.448096 | -0.000010 | -0.000559 | -0.000019 | -0.000558 |
| 26296 | 1.724934 | 1.735250 | -1.500642 | 1.260881 | 0.775294 | 0.854617 | 0.982577 | 2.458770 | -0.707107 | -0.707107 | ... | -0.000033 | 1.412765 | -0.000010 | -0.000559 | -0.000019 | -0.000558 | -0.000010 | 2.448094 | -0.000019 | 2.448096 |
| 26297 | 1.724934 | 1.735250 | -1.500642 | 1.260881 | 0.775294 | 0.854617 | 0.982577 | 2.458770 | -0.707107 | -0.707107 | ... | -0.000033 | 1.412765 | -0.000010 | -0.000559 | -0.000019 | -0.000558 | -0.000010 | -0.000559 | -0.000019 | -0.000558 |
26298 rows × 70 columns
linear_pipe = Pipeline(steps = [
('prep', linear_prep),
('model', Ridge())
])
linear_pipe.fit(X, log_y)
linear_pipe[0].transform(X_test)
| day_of_year | day_of_month | day_of_week | log_gdp | days_from_easter | days_from_mom | days_from_wed | days_from_sun | country_Finland | country_Norway | ... | day_of_year_sin_2 | day_of_year_cos_2 | day_of_year_sin_1_mul_product_Kaggle Mug | day_of_year_cos_1_mul_product_Kaggle Mug | day_of_year_sin_2_mul_product_Kaggle Mug | day_of_year_cos_2_mul_product_Kaggle Mug | day_of_year_sin_1_mul_product_Kaggle Hat | day_of_year_cos_1_mul_product_Kaggle Hat | day_of_year_sin_2_mul_product_Kaggle Hat | day_of_year_cos_2_mul_product_Kaggle Hat | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| row_id | |||||||||||||||||||||
| 26298 | -1.727310 | -1.673805 | -1.00077 | -1.065656 | -1.503789 | -1.192261 | -1.036474 | -0.421954 | 1.414214 | -0.707107 | ... | 0.048663 | 1.411927 | 0.042169 | 2.447731 | 0.084325 | 2.446645 | -0.000010 | -0.000559 | -0.000019 | -0.000558 |
| 26299 | -1.727310 | -1.673805 | -1.00077 | -1.065656 | -1.503789 | -1.192261 | -1.036474 | -0.421954 | 1.414214 | -0.707107 | ... | 0.048663 | 1.411927 | -0.000010 | -0.000559 | -0.000019 | -0.000558 | 0.042169 | 2.447731 | 0.084325 | 2.446645 |
| 26300 | -1.727310 | -1.673805 | -1.00077 | -1.065656 | -1.503789 | -1.192261 | -1.036474 | -0.421954 | 1.414214 | -0.707107 | ... | 0.048663 | 1.411927 | -0.000010 | -0.000559 | -0.000019 | -0.000558 | -0.000010 | -0.000559 | -0.000019 | -0.000558 |
| 26301 | -1.727310 | -1.673805 | -1.00077 | -1.065656 | -1.503789 | -1.192261 | -1.036474 | -0.421954 | 1.414214 | -0.707107 | ... | 0.048663 | 1.411927 | 0.042169 | 2.447731 | 0.084325 | 2.446645 | -0.000010 | -0.000559 | -0.000019 | -0.000558 |
| 26302 | -1.727310 | -1.673805 | -1.00077 | -1.065656 | -1.503789 | -1.192261 | -1.036474 | -0.421954 | 1.414214 | -0.707107 | ... | 0.048663 | 1.411927 | -0.000010 | -0.000559 | -0.000019 | -0.000558 | 0.042169 | 2.447731 | 0.084325 | 2.446645 |
| ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... |
| 32863 | 1.724934 | 1.735250 | -1.00077 | 1.133907 | 0.775294 | 0.854617 | 0.982577 | 2.458770 | -0.707107 | -0.707107 | ... | -0.000033 | 1.412765 | -0.000010 | -0.000559 | -0.000019 | -0.000558 | -0.000010 | 2.448094 | -0.000019 | 2.448096 |
| 32864 | 1.724934 | 1.735250 | -1.00077 | 1.133907 | 0.775294 | 0.854617 | 0.982577 | 2.458770 | -0.707107 | -0.707107 | ... | -0.000033 | 1.412765 | -0.000010 | -0.000559 | -0.000019 | -0.000558 | -0.000010 | -0.000559 | -0.000019 | -0.000558 |
| 32865 | 1.724934 | 1.735250 | -1.00077 | 1.133907 | 0.775294 | 0.854617 | 0.982577 | 2.458770 | -0.707107 | -0.707107 | ... | -0.000033 | 1.412765 | -0.000010 | 2.448094 | -0.000019 | 2.448096 | -0.000010 | -0.000559 | -0.000019 | -0.000558 |
| 32866 | 1.724934 | 1.735250 | -1.00077 | 1.133907 | 0.775294 | 0.854617 | 0.982577 | 2.458770 | -0.707107 | -0.707107 | ... | -0.000033 | 1.412765 | -0.000010 | -0.000559 | -0.000019 | -0.000558 | -0.000010 | 2.448094 | -0.000019 | 2.448096 |
| 32867 | 1.724934 | 1.735250 | -1.00077 | 1.133907 | 0.775294 | 0.854617 | 0.982577 | 2.458770 | -0.707107 | -0.707107 | ... | -0.000033 | 1.412765 | -0.000010 | -0.000559 | -0.000019 | -0.000558 | -0.000010 | -0.000559 | -0.000019 | -0.000558 |
6570 rows × 70 columns
linear_pred_train = np.exp(linear_pipe.predict(X))
linear_pred_test = np.exp(linear_pipe.predict(X_test))
def plot_oof(y, y_pred):
plt.figure(figsize=(10, 10))
plt.scatter(y, y_pred, s=1, color='r')
#plt.scatter(np.log(y_va), np.log(y_va_pred), s=1, color='g')
plt.plot([plt.xlim()[0], plt.xlim()[1]], [plt.xlim()[0], plt.xlim()[1]], '--', color='k')
plt.gca().set_aspect('equal')
plt.xlabel('y_true')
plt.ylabel('y_pred')
plt.title('OOF Predictions')
plt.show()
plot_oof(y, linear_pred_train)
def plot_data(df, y_pred, country = 'Norway', store = 'KaggleMart', product = 'Kaggle Hat', figsize = (20,6)):
plot_data = df.query(f'country == "{country}" and store == "{store}" and product == "{product}"')
idx = plot_data.index
plt.figure(figsize=figsize)
plt.plot(plot_data.date, y_pred[idx], label='prediction')
plt.scatter(plot_data.date, plot_data.num_sold, label='true', alpha=0.5, color='red', s=3)
plt.legend()
plot_data(df_train, linear_pred_train)
df_train['pred'] = linear_pred_train
by_date = df_train.groupby('date')
residuals = by_date.pred.mean() - by_date.num_sold.mean()
plt.figure(figsize = (20, 8))
plt.plot(df_train.date, df_train.num_sold, alpha = 0.2, c = 'r')
plt.scatter(df_train.date.unique(), residuals, s = 3)
<matplotlib.collections.PathCollection at 0x7effcc508d10>
y_resid = df_train.num_sold - linear_pred_train
y_resid
row_id
0 1.600911
1 12.788551
2 1.253354
3 1.103563
4 26.558905
...
26293 -21.426569
26294 8.077731
26295 49.874044
26296 -31.454064
26297 -33.847726
Name: num_sold, Length: 26298, dtype: float64
boosting_prep = Pipeline(steps = [
('cat_imp', CategoricalImputer()),
('OrdinalEncoding', OrdinalEncoder(encoding_method='ordered')),
('cyc1', cyc_1),
('cyc2', cyc_2),
# ('combo', CombineWithReferenceFeature(variables_to_combine=['day_of_year_sin_1', 'day_of_year_cos_1',
# 'day_of_year_sin_2', 'day_of_year_cos_2'],
# reference_variables=['product_Kaggle Mug', 'product_Kaggle Hat'],
# operations = ['mul'])),
('drop', DropFeatures(features_to_drop=['day_of_year','day_of_week','day_of_month','year',]
# 'store_KaggleMart','product_Kaggle Sticker', 'week_week','country_Sweden']
)),
])
boosting_prep.fit_transform(X,y)
| country | store | product | log_gdp | week | finland_holiday | norway_holiday | sweden_holiday | days_from_easter | days_from_mom | days_from_wed | days_from_sun | day_of_year_sin_1 | day_of_year_cos_1 | day_of_year_sin_2 | day_of_year_cos_2 | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| row_id | ||||||||||||||||
| 0 | 0 | 0 | 1 | 5.457200 | 0 | 15 | 0 | 7 | -5 | -1 | -5 | -1 | 1.721336e-02 | 0.999852 | 3.442161e-02 | 0.999407 |
| 1 | 0 | 0 | 2 | 5.457200 | 0 | 15 | 0 | 7 | -5 | -1 | -5 | -1 | 1.721336e-02 | 0.999852 | 3.442161e-02 | 0.999407 |
| 2 | 0 | 0 | 0 | 5.457200 | 0 | 15 | 0 | 7 | -5 | -1 | -5 | -1 | 1.721336e-02 | 0.999852 | 3.442161e-02 | 0.999407 |
| 3 | 0 | 1 | 1 | 5.457200 | 0 | 15 | 0 | 7 | -5 | -1 | -5 | -1 | 1.721336e-02 | 0.999852 | 3.442161e-02 | 0.999407 |
| 4 | 0 | 1 | 2 | 5.457200 | 0 | 15 | 0 | 7 | -5 | -1 | -5 | -1 | 1.721336e-02 | 0.999852 | 3.442161e-02 | 0.999407 |
| ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... |
| 26293 | 1 | 0 | 2 | 6.319788 | 0 | 12 | 0 | 16 | 65 | 9 | 5 | 9 | -2.449294e-16 | 1.000000 | -4.898587e-16 | 1.000000 |
| 26294 | 1 | 0 | 0 | 6.319788 | 0 | 12 | 0 | 16 | 65 | 9 | 5 | 9 | -2.449294e-16 | 1.000000 | -4.898587e-16 | 1.000000 |
| 26295 | 1 | 1 | 1 | 6.319788 | 0 | 12 | 0 | 16 | 65 | 9 | 5 | 9 | -2.449294e-16 | 1.000000 | -4.898587e-16 | 1.000000 |
| 26296 | 1 | 1 | 2 | 6.319788 | 0 | 12 | 0 | 16 | 65 | 9 | 5 | 9 | -2.449294e-16 | 1.000000 | -4.898587e-16 | 1.000000 |
| 26297 | 1 | 1 | 0 | 6.319788 | 0 | 12 | 0 | 16 | 65 | 9 | 5 | 9 | -2.449294e-16 | 1.000000 | -4.898587e-16 | 1.000000 |
26298 rows × 16 columns
from xgboost import XGBRegressor
boosting_pipe = Pipeline(steps = [
('prep', boosting_prep),
('model', XGBRegressor())
])
boosting_pipe.fit(X, y_resid)
boosting_pred_train = boosting_pipe.predict(X)
boosting_pred_test = boosting_pipe.predict(X_test)
plot_data(df_train, boosting_pred_train)
plt.plot(df_test.date, boosting_pred_test)
[<matplotlib.lines.Line2D at 0x7effcc06fed0>]
plot_oof(y, boosting_pred_train)
y_pred_train = linear_pred_train + boosting_pred_train
y_pred_test = linear_pred_test + boosting_pred_test
plot_oof(y, y_pred_train)
plot_data(df_train, y_pred_train, country = 'Finland', store = 'KaggleRama', product = 'Kaggle Hat')
def plot_test(df, y_pred, country = 'Finland', store = 'KaggleMart', product = 'Kaggle Hat', figsize = (20,6)):
plot_data = df.reset_index(drop = True)
plot_data = plot_data.query(f'country == "{country}" and store == "{store}" and product == "{product}"')
idx = plot_data.index
y_pred = pd.Series(y_pred)
y_pred.iloc[idx].plot(figsize = figsize)
plot_test(df_test, y_pred_test)
y_pred_test = pd.Series(y_pred_test)
y_pred_test.index = df_test.index
y_pred_test.index.name = 'row_id'
y_pred_test.name = 'num_sold'
y_pred_test.to_csv('submission_hybrid_v2.csv')
y_pred_train
array([ 333.2296468 , 520.56556714, 152.48215013, ..., 978.08642442,
1464.8970567 , 409.52754557])
y_pred_test
row_id
26298 384.855167
26299 604.195058
26300 172.140650
26301 682.554481
26302 1029.610388
...
32863 794.426730
32864 228.970733
32865 934.668146
32866 1392.851176
32867 409.546574
Name: num_sold, Length: 6570, dtype: float64
from sklearn.linear_model import Ridge
from xgboost import XGBRegressor
import data as d
from models import HybridModel
from preprocessing import preprocess_dict
from utils import Preprocess
GDP_PATH = '../inputs/GDP_data_2015_to_2019_Finland_Norway_Sweden.csv'
TRAIN_PATH = '../inputs/train.csv'
TEST_PATH = '../inputs/test.csv'
feature_engineer = Preprocess(GDP_PATH, d.EASTER_DICT, d.SPECIAL_DAYS_DICT, d.COUNTRY_HOLIDAYS_DICT, )
train_df = feature_engineer(TRAIN_PATH)
test_df = feature_engineer(TEST_PATH, name = 'Test')
print(train_df)
X = train_df.drop(columns = ['gdp','date','num_sold'])
y = train_df.num_sold
X_test = test_df.drop(columns = ['gdp','date'])
linear_prep = preprocess_dict['linear_v1']
boosting_prep = preprocess_dict['boosting_v1']
X_l = linear_prep.fit_transform(X, y)
X_l = linear_prep.transform(X_test, y)
X_b = boosting_prep.fit_transform(X, y)
model = HybridModel(Ridge(), XGBRegressor())
model.fit(X_l, X_b, y)
print(X_l)
print(X_b)
print(model.predict(X_l, X_b))
print(model.predict(X_l, X_b))
Train Set created with 19 features
Test Set created with 18 features
date country store product num_sold day_of_year \
row_id
0 2015-01-01 Finland KaggleMart Kaggle Mug 329 1
1 2015-01-01 Finland KaggleMart Kaggle Hat 520 1
2 2015-01-01 Finland KaggleMart Kaggle Sticker 146 1
3 2015-01-01 Finland KaggleRama Kaggle Mug 572 1
4 2015-01-01 Finland KaggleRama Kaggle Hat 911 1
... ... ... ... ... ... ...
26293 2018-12-31 Sweden KaggleMart Kaggle Hat 823 365
26294 2018-12-31 Sweden KaggleMart Kaggle Sticker 250 365
26295 2018-12-31 Sweden KaggleRama Kaggle Mug 1004 365
26296 2018-12-31 Sweden KaggleRama Kaggle Hat 1441 365
26297 2018-12-31 Sweden KaggleRama Kaggle Sticker 388 365
day_of_month day_of_week year gdp log_gdp week \
row_id
0 1 3 2015 234.440 5.457200 week
1 1 3 2015 234.440 5.457200 week
2 1 3 2015 234.440 5.457200 week
3 1 3 2015 234.440 5.457200 week
4 1 3 2015 234.440 5.457200 week
... ... ... ... ... ... ...
26293 31 0 2018 555.455 6.319788 week
26294 31 0 2018 555.455 6.319788 week
26295 31 0 2018 555.455 6.319788 week
26296 31 0 2018 555.455 6.319788 week
26297 31 0 2018 555.455 6.319788 week
finland_holiday norway_holiday sweden_holiday days_from_easter \
row_id
0 Uudenvuodenpäivä NaN NaN -5
1 Uudenvuodenpäivä NaN NaN -5
2 Uudenvuodenpäivä NaN NaN -5
3 Uudenvuodenpäivä NaN NaN -5
4 Uudenvuodenpäivä NaN NaN -5
... ... ... ... ...
26293 NaN NaN Nyårsafton 65
26294 NaN NaN Nyårsafton 65
26295 NaN NaN Nyårsafton 65
26296 NaN NaN Nyårsafton 65
26297 NaN NaN Nyårsafton 65
days_from_mom days_from_wed days_from_sun
row_id
0 -1 -5 -1
1 -1 -5 -1
2 -1 -5 -1
3 -1 -5 -1
4 -1 -5 -1
... ... ... ...
26293 9 5 9
26294 9 5 9
26295 9 5 9
26296 9 5 9
26297 9 5 9
[26298 rows x 19 columns]
y_resid: [ 1.60091054 12.78855121 1.25335354 ... 49.87404387 -31.45406353
-33.84772551]
day_of_year day_of_month day_of_week log_gdp days_from_easter \
row_id
0 -1.727310 -1.673805 -0.001026 -1.503796 -1.503789
1 -1.727310 -1.673805 -0.001026 -1.503796 -1.503789
2 -1.727310 -1.673805 -0.001026 -1.503796 -1.503789
3 -1.727310 -1.673805 -0.001026 -1.503796 -1.503789
4 -1.727310 -1.673805 -0.001026 -1.503796 -1.503789
... ... ... ... ... ...
26293 1.724934 1.735250 -1.500642 1.260881 0.775294
26294 1.724934 1.735250 -1.500642 1.260881 0.775294
26295 1.724934 1.735250 -1.500642 1.260881 0.775294
26296 1.724934 1.735250 -1.500642 1.260881 0.775294
26297 1.724934 1.735250 -1.500642 1.260881 0.775294
days_from_mom days_from_wed days_from_sun country_Finland \
row_id
0 -1.192261 -1.036474 -0.421954 1.414214
1 -1.192261 -1.036474 -0.421954 1.414214
2 -1.192261 -1.036474 -0.421954 1.414214
3 -1.192261 -1.036474 -0.421954 1.414214
4 -1.192261 -1.036474 -0.421954 1.414214
... ... ... ... ...
26293 0.854617 0.982577 2.458770 -0.707107
26294 0.854617 0.982577 2.458770 -0.707107
26295 0.854617 0.982577 2.458770 -0.707107
26296 0.854617 0.982577 2.458770 -0.707107
26297 0.854617 0.982577 2.458770 -0.707107
country_Norway ... day_of_year_sin_2 day_of_year_cos_2 \
row_id ...
0 -0.707107 ... 0.048663 1.411927
1 -0.707107 ... 0.048663 1.411927
2 -0.707107 ... 0.048663 1.411927
3 -0.707107 ... 0.048663 1.411927
4 -0.707107 ... 0.048663 1.411927
... ... ... ... ...
26293 -0.707107 ... -0.000033 1.412765
26294 -0.707107 ... -0.000033 1.412765
26295 -0.707107 ... -0.000033 1.412765
26296 -0.707107 ... -0.000033 1.412765
26297 -0.707107 ... -0.000033 1.412765
day_of_year_sin_1_mul_product_Kaggle Mug \
row_id
0 0.042169
1 -0.000010
2 -0.000010
3 0.042169
4 -0.000010
... ...
26293 -0.000010
26294 -0.000010
26295 -0.000010
26296 -0.000010
26297 -0.000010
day_of_year_cos_1_mul_product_Kaggle Mug \
row_id
0 2.447731
1 -0.000559
2 -0.000559
3 2.447731
4 -0.000559
... ...
26293 -0.000559
26294 -0.000559
26295 2.448094
26296 -0.000559
26297 -0.000559
day_of_year_sin_2_mul_product_Kaggle Mug \
row_id
0 0.084325
1 -0.000019
2 -0.000019
3 0.084325
4 -0.000019
... ...
26293 -0.000019
26294 -0.000019
26295 -0.000019
26296 -0.000019
26297 -0.000019
day_of_year_cos_2_mul_product_Kaggle Mug \
row_id
0 2.446645
1 -0.000558
2 -0.000558
3 2.446645
4 -0.000558
... ...
26293 -0.000558
26294 -0.000558
26295 2.448096
26296 -0.000558
26297 -0.000558
day_of_year_sin_1_mul_product_Kaggle Hat \
row_id
0 -0.000010
1 0.042169
2 -0.000010
3 -0.000010
4 0.042169
... ...
26293 -0.000010
26294 -0.000010
26295 -0.000010
26296 -0.000010
26297 -0.000010
day_of_year_cos_1_mul_product_Kaggle Hat \
row_id
0 -0.000559
1 2.447731
2 -0.000559
3 -0.000559
4 2.447731
... ...
26293 2.448094
26294 -0.000559
26295 -0.000559
26296 2.448094
26297 -0.000559
day_of_year_sin_2_mul_product_Kaggle Hat \
row_id
0 -0.000019
1 0.084325
2 -0.000019
3 -0.000019
4 0.084325
... ...
26293 -0.000019
26294 -0.000019
26295 -0.000019
26296 -0.000019
26297 -0.000019
day_of_year_cos_2_mul_product_Kaggle Hat
row_id
0 -0.000558
1 2.446645
2 -0.000558
3 -0.000558
4 2.446645
... ...
26293 2.448096
26294 -0.000558
26295 -0.000558
26296 2.448096
26297 -0.000558
[26298 rows x 70 columns]
country store product log_gdp week finland_holiday \
row_id
0 0 0 1 5.457200 0 15
1 0 0 2 5.457200 0 15
2 0 0 0 5.457200 0 15
3 0 1 1 5.457200 0 15
4 0 1 2 5.457200 0 15
... ... ... ... ... ... ...
26293 1 0 2 6.319788 0 12
26294 1 0 0 6.319788 0 12
26295 1 1 1 6.319788 0 12
26296 1 1 2 6.319788 0 12
26297 1 1 0 6.319788 0 12
norway_holiday sweden_holiday days_from_easter days_from_mom \
row_id
0 0 7 -5 -1
1 0 7 -5 -1
2 0 7 -5 -1
3 0 7 -5 -1
4 0 7 -5 -1
... ... ... ... ...
26293 0 16 65 9
26294 0 16 65 9
26295 0 16 65 9
26296 0 16 65 9
26297 0 16 65 9
days_from_wed days_from_sun day_of_year_sin_1 day_of_year_cos_1 \
row_id
0 -5 -1 1.721336e-02 0.999852
1 -5 -1 1.721336e-02 0.999852
2 -5 -1 1.721336e-02 0.999852
3 -5 -1 1.721336e-02 0.999852
4 -5 -1 1.721336e-02 0.999852
... ... ... ... ...
26293 5 9 -2.449294e-16 1.000000
26294 5 9 -2.449294e-16 1.000000
26295 5 9 -2.449294e-16 1.000000
26296 5 9 -2.449294e-16 1.000000
26297 5 9 -2.449294e-16 1.000000
day_of_year_sin_2 day_of_year_cos_2
row_id
0 3.442161e-02 0.999407
1 3.442161e-02 0.999407
2 3.442161e-02 0.999407
3 3.442161e-02 0.999407
4 3.442161e-02 0.999407
... ... ...
26293 -4.898587e-16 1.000000
26294 -4.898587e-16 1.000000
26295 -4.898587e-16 1.000000
26296 -4.898587e-16 1.000000
26297 -4.898587e-16 1.000000
[26298 rows x 16 columns]
[ 331.63600218 509.84360564 150.36060196 ... 988.92224748 1462.16449559
413.80443633]
[ 331.63600218 509.84360564 150.36060196 ... 988.92224748 1462.16449559
413.80443633]
boosting_prep
Pipeline(steps=[('cat_imp', CategoricalImputer()), ('oe', OrdinalEncoder()),
('cyc1', CyclicalTransformerV2(suffix='_1')),
('cyc2', CyclicalTransformerV2(suffix='_2')),
('drop',
DropFeatures(features_to_drop=['day_of_year', 'day_of_week',
'day_of_month', 'year']))])
df = pd.DataFrame()
df.loc[2, 'KFold'] = 1
df.loc[1, 'KFold'] = 0
df.loc[3:6, 'KFold'] = 3
df
| KFold | |
|---|---|
| 2 | 1.0 |
| 1 | 0.0 |
import pandas as pd
import numpy as np
pd.DataFrame(np.nan, index = range(5), columns = ['fold','num_sold'])
| fold | num_sold | |
|---|---|---|
| 0 | NaN | NaN |
| 1 | NaN | NaN |
| 2 | NaN | NaN |
| 3 | NaN | NaN |
| 4 | NaN | NaN |