light GBM

https://www.kaggle.com/code/dangnguyen97/0-38006-lightgbm

라이브러리 임포트

import os
                        
import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
import seaborn as sns
​
from darts import TimeSeries
from darts.dataprocessing import Pipeline
from darts.dataprocessing.transformers import Scaler, InvertibleMapper, StaticCovariatesTransformer
from darts.dataprocessing.transformers.missing_values_filler import MissingValuesFiller
from darts.metrics import rmsle
from darts.models import LinearRegressionModel, LightGBMModel, XGBModel, CatBoostModel
from darts.models.filtering.moving_average_filter import MovingAverageFilter
from sklearn.preprocessing import MinMaxScaler, OneHotEncoder
from tqdm.notebook import tqdm_notebook
​
plt.style.use("ggplot")
plt.rcParams["font.size"] = 15
COLORS = list(sns.color_palette())

 

내용출력함수 정의

# helper function to print messages
def cprint(title, *args):
    print(
        "="*len(title), title, "="*len(title),
        *args,
        sep="\n",
    )

데이터 받기

PATH = "/kaggle/input/store-sales-time-series-forecasting"

train = pd.read_csv(os.path.join(PATH, "train.csv"), parse_dates=["date"])
test = pd.read_csv(os.path.join(PATH, "test.csv"), parse_dates=["date"])

oil = pd.read_csv(os.path.join(PATH, "oil.csv"), parse_dates=["date"]).rename(columns={"dcoilwtico": "oil"})
store = pd.read_csv(os.path.join(PATH, "stores.csv"))
transaction = pd.read_csv(os.path.join(PATH, "transactions.csv"), parse_dates=["date"])
holiday = pd.read_csv(os.path.join(PATH, "holidays_events.csv"), parse_dates=["date"])

train.head(5)

데이터 분석 출력

num_family = train.family.nunique()
num_store = train.store_nbr.nunique()
num_ts = train.groupby(["store_nbr", "family"]).ngroups
train_start = train.date.min().date()
train_end = train.date.max().date()
num_train_date = train.date.nunique()
train_len = (train_end - train_start).days + 1
test_start = test.date.min().date()
test_end = test.date.max().date()
num_test_date = test.date.nunique()
test_len = (test_end - test_start).days + 1

 

cprint(
    "Basic information of data",
    f"Number of family types      : {num_family}",
    f"Number of stores            : {num_store}",
    f"Number of store-family pairs: {num_family * num_store}",
    f"Number of target series     : {num_ts}",
    "",
    f"Number of unique train dates: {num_train_date}",
    f"Train date range            : {train_len} days from {train_start} to {train_end}",
    f"Number of unique test dates : {num_test_date}",
    f"Test date range             : {test_len} days from {test_start} to {test_end}",
)

=========================
Basic information of data
=========================
Number of family types      : 33
Number of stores            : 54
Number of store-family pairs: 1782
Number of target series     : 1782

Number of unique train dates: 1684
Train date range            : 1688 days from 2013-01-01 to 2017-08-15
Number of unique test dates : 16
Test date range             : 16 days from 2017-08-16 to 2017-08-31