/ EXERCISE

Multiple Time Series Forecasting with PyCaret

Multiple Time Series Forecasting with PyCaret

Imports

import pandas as pd
import numpy as np
import warnings
warnings.filterwarnings('ignore')

import plotly.express as px
from pycaret.regression import *
from tqdm import tqdm

data = pd.read_csv('../exercise/store_item/train.csv')
data['date'] = pd.to_datetime(data['date'])

data['store'] = ['store_' + str(i) for i in data['store']]
data['item'] = ['item_' + str(i) for i in data['item']]
data['time_series'] = data[['store', 'item']].apply(lambda x: '_'.join(x), axis=1)
data.drop(['store', 'item'], axis=1, inplace=True)

data['month'] = [i.month for i in data['date']]
data['year'] = [i.year for i in data['date']]
data['day_of_week'] = [i.dayofweek for i in data['date']]
data['day_of_year'] = [i.dayofyear for i in data['date']]

data.head()
date sales time_series month year day_of_week day_of_year
0 2013-01-01 13 store_1_item_1 1 2013 1 1
1 2013-01-02 11 store_1_item_1 1 2013 2 2
2 2013-01-03 14 store_1_item_1 1 2013 3 3
3 2013-01-04 13 store_1_item_1 1 2013 4 4
4 2013-01-05 10 store_1_item_1 1 2013 5 5 
data['time_series'].nunique()

500

Plot

idx = data['time_series'].unique()[0]
subset = data[data['time_series'] == idx]
subset['MA'] = subset['sales'].rolling(30).mean()
fig = px.line(subset, x='date', y=['sales', 'MA'], title=idx, template='plotly_dark')
fig.show()

image1

Training for store_1_item_1

df_subset = data[data['time_series'] == idx]

df_subset.shape

(1826, 7)

s = setup(df_subset, target='sales', train_size=0.95, data_split_shuffle=False, fold_strategy='timeseries', fold=3, 
         ignore_features = ['date', 'time_series'], numeric_features=['day_of_year','year'],
          categorical_features=['month', 'day_of_week'], session_id = 1234)
Description Value
0 session_id 1234
1 Target sales
2 Original Data (1826, 7)
3 Missing Values False
4 Numeric Features 2
5 Categorical Features 2
6 Ordinal Features False
7 High Cardinality Features False
8 High Cardinality Method None
9 Transformed Train Set (1734, 21)
10 Transformed Test Set (92, 21)
11 Shuffle Train-Test False
12 Stratify Train-Test False
13 Fold Generator TimeSeriesSplit
14 Fold Number 3
15 CPU Jobs -1
16 Use GPU False
17 Log Experiment False
18 Experiment Name reg-default-name
19 USI cd06
20 Imputation Type simple
21 Iterative Imputation Iteration None
22 Numeric Imputer mean
23 Iterative Imputation Numeric Model None
24 Categorical Imputer constant
25 Iterative Imputation Categorical Model None
26 Unknown Categoricals Handling least_frequent
27 Normalize False
28 Normalize Method None
29 Transformation False
30 Transformation Method None
31 PCA False
32 PCA Method None
33 PCA Components None
34 Ignore Low Variance False
35 Combine Rare Levels False
36 Rare Level Threshold None
37 Numeric Binning False
38 Remove Outliers False
39 Outliers Threshold None
40 Remove Multicollinearity False
41 Multicollinearity Threshold None
42 Remove Perfect Collinearity True
43 Clustering False
44 Clustering Iteration None
45 Polynomial Features False
46 Polynomial Degree None
47 Trignometry Features False
48 Polynomial Threshold None
49 Group Features False
50 Feature Selection False
51 Feature Selection Method classic
52 Features Selection Threshold None
53 Feature Interaction False
54 Feature Ratio False
55 Interaction Threshold None
56 Transform Target False
57 Transform Target Method box-cox 
best_model = compare_models(sort='MAE')
Model MAE MSE RMSE R2 RMSLE MAPE TT (Sec)
br Bayesian Ridge 3.7656 22.3829 4.7245 0.4782 0.2346 0.2114 0.0067
ridge Ridge Regression 3.7711 22.4091 4.7278 0.4772 0.2352 0.2124 0.5100
lr Linear Regression 3.7802 22.4823 4.7364 0.4747 0.2360 0.2137 0.7267
lightgbm Light Gradient Boosting Machine 3.9912 25.6055 5.0594 0.3971 0.2429 0.2038 0.0333
gbr Gradient Boosting Regressor 4.0101 26.1919 5.1171 0.3830 0.2422 0.2001 0.0333
ada AdaBoost Regressor 4.1772 28.3098 5.3089 0.3409 0.2481 0.2129 0.0433
rf Random Forest Regressor 4.2648 29.6002 5.4396 0.3006 0.2641 0.2153 0.2800
huber Huber Regressor 4.3631 31.2640 5.5737 0.2780 0.2552 0.2028 0.0200
xgboost Extreme Gradient Boosting 4.5708 33.5601 5.7889 0.2151 0.2863 0.2276 0.2400
et Extra Trees Regressor 4.5897 34.1530 5.8431 0.1910 0.2934 0.2352 0.2400
knn K Neighbors Regressor 4.8785 37.8773 6.1444 0.1191 0.2891 0.2424 0.0467
dt Decision Tree Regressor 5.1239 42.4573 6.5113 0.0035 0.3303 0.2628 0.0100
omp Orthogonal Matching Pursuit 5.3050 45.5557 6.7185 -0.0503 0.3137 0.2706 0.0067
en Elastic Net 5.5900 50.1346 7.0523 -0.1647 0.3285 0.2809 0.5267
lasso Lasso Regression 5.6448 51.1533 7.1169 -0.1856 0.3314 0.2824 0.5167
llar Lasso Least Angle Regression 5.8939 56.1759 7.4654 -0.2946 0.3417 0.2815 0.0067
par Passive Aggressive Regressor 6.3936 64.9179 8.0232 -0.5318 0.3753 0.2990 0.0067
lar Least Angle Regression 1084.2732 4596012.7859 1251.2497 -100016.0471 1.9992 53.5604 0.0100 
f = finalize_model(best_model)

Generate predictions

all_dates = pd.date_range(start='2013-01-01', end='2019-12-31', freq='D')

score_df = pd.DataFrame()
score_df['date'] = all_dates
score_df['month'] = [i.month for i in score_df['date']]
score_df['year'] = [i.year for i in score_df['date']]
score_df['day_of_week'] = [i.dayofweek for i in score_df['date']]
score_df['day_of_year'] = [i.dayofyear for i in score_df['date']]

score_df
date month year day_of_week day_of_year
0 2013-01-01 1 2013 1 1
1 2013-01-02 1 2013 2 2
2 2013-01-03 1 2013 3 3
3 2013-01-04 1 2013 4 4
4 2013-01-05 1 2013 5 5
... ... ... ... ... ...
2551 2019-12-27 12 2019 4 361
2552 2019-12-28 12 2019 5 362
2553 2019-12-29 12 2019 6 363
2554 2019-12-30 12 2019 0 364
2555 2019-12-31 12 2019 1 365

2556 rows × 5 columns

from pycaret.regression import predict_model

score = []

p = predict_model(f, data=score_df)

p['time_series'] = idx

p
date month year day_of_week day_of_year Label time_series
0 2013-01-01 1 2013 1 1 8.927565 store_1_item_1
1 2013-01-02 1 2013 2 2 9.515175 store_1_item_1
2 2013-01-03 1 2013 3 3 10.213244 store_1_item_1
3 2013-01-04 1 2013 4 4 11.780002 store_1_item_1
4 2013-01-05 1 2013 5 5 13.710251 store_1_item_1
... ... ... ... ... ... ... ...
2551 2019-12-27 12 2019 4 361 21.706059 store_1_item_1
2552 2019-12-28 12 2019 5 362 23.636309 store_1_item_1
2553 2019-12-29 12 2019 6 363 24.467096 store_1_item_1
2554 2019-12-30 12 2019 0 364 16.320039 store_1_item_1
2555 2019-12-31 12 2019 1 365 18.949996 store_1_item_1

2556 rows × 7 columns

final_df = pd.merge(p, data, how='left', left_on = ['date','time_series'], right_on=['date','time_series'])

final_df
date month_x year_x day_of_week_x day_of_year_x Label time_series sales month_y year_y day_of_week_y day_of_year_y
0 2013-01-01 1 2013 1 1 8.927565 store_1_item_1 13.0 1.0 2013.0 1.0 1.0
1 2013-01-02 1 2013 2 2 9.515175 store_1_item_1 11.0 1.0 2013.0 2.0 2.0
2 2013-01-03 1 2013 3 3 10.213244 store_1_item_1 14.0 1.0 2013.0 3.0 3.0
3 2013-01-04 1 2013 4 4 11.780002 store_1_item_1 13.0 1.0 2013.0 4.0 4.0
4 2013-01-05 1 2013 5 5 13.710251 store_1_item_1 10.0 1.0 2013.0 5.0 5.0
... ... ... ... ... ... ... ... ... ... ... ... ...
2551 2019-12-27 12 2019 4 361 21.706059 store_1_item_1 NaN NaN NaN NaN NaN
2552 2019-12-28 12 2019 5 362 23.636309 store_1_item_1 NaN NaN NaN NaN NaN
2553 2019-12-29 12 2019 6 363 24.467096 store_1_item_1 NaN NaN NaN NaN NaN
2554 2019-12-30 12 2019 0 364 16.320039 store_1_item_1 NaN NaN NaN NaN NaN
2555 2019-12-31 12 2019 1 365 18.949996 store_1_item_1 NaN NaN NaN NaN NaN

2556 rows × 12 columns

fig = px.line(final_df, x='date',y=['sales','Label'], title=idx, template='plotly_dark')
fig.show()

image2

Reference : Multiple Time Series Forecasting with PyCaret