728x90

https://dacon.io/competitions/open/235536/overview/description

[문화] 영화 관객수 예측 모델 개발 - DACON

좋아요는 1분 내에 한 번만 클릭 할 수 있습니다.

dacon.io

영화 관객 수 데이터를 활용한 데이터 분석¶

1. 라이브러리 및 데이터¶

Library & Data¶

In [3]:

# 패키지 불러오기

import pandas as pd
import lightgbm as lgb

In [11]:

# train, test, submission dara 불러오기

train = pd.read_csv('data/movies_train.csv')
test = pd.read_csv('data/movies_test.csv')
submission = pd.read_csv('data/submission.csv')

2. 탐색적 자료분석¶

Exploratory Data Analysis (EDA)¶

In [12]:

# train data 상위 5개의 행 출력
train.head()

Out[12]:

	title	distributor	genre	release_time	time	screening_rat	director	dir_prev_bfnum	dir_prev_num	num_staff	num_actor	box_off_num
0	개들의 전쟁	롯데엔터테인먼트	액션	2012-11-22	96	청소년 관람불가	조병옥	NaN	0	91	2	23398
1	내부자들	(주)쇼박스	느와르	2015-11-19	130	청소년 관람불가	우민호	1161602.50	2	387	3	7072501
2	은밀하게 위대하게	(주)쇼박스	액션	2013-06-05	123	15세 관람가	장철수	220775.25	4	343	4	6959083
3	나는 공무원이다	(주)NEW	코미디	2012-07-12	101	전체 관람가	구자홍	23894.00	2	20	6	217866
4	불량남녀	쇼박스(주)미디어플렉스	코미디	2010-11-04	108	15세 관람가	신근호	1.00	1	251	2	483387

In [13]:

# test data 상위 5개의 행 출력

test.head()

Out[13]:

	title	distributor	genre	release_time	time	screening_rat	director	dir_prev_bfnum	dir_prev_num	num_staff	num_actor
0	용서는 없다	시네마서비스	느와르	2010-01-07	125	청소년 관람불가	김형준	3.005290e+05	2	304	3
1	아빠가 여자를 좋아해	(주)쇼박스	멜로/로맨스	2010-01-14	113	12세 관람가	이광재	3.427002e+05	4	275	3
2	하모니	CJ 엔터테인먼트	드라마	2010-01-28	115	12세 관람가	강대규	4.206611e+06	3	419	7
3	의형제	(주)쇼박스	액션	2010-02-04	116	15세 관람가	장훈	6.913420e+05	2	408	2
4	평행 이론	CJ 엔터테인먼트	공포	2010-02-18	110	15세 관람가	권호영	3.173800e+04	1	380	1

In [14]:

# submission data 상위 5개의 행 출력

submission.head()

Out[14]:

	title	box_off_num
0	용서는 없다	0
1	아빠가 여자를 좋아해	0
2	하모니	0
3	의형제	0
4	평행 이론	0

In [15]:

# train data 하위 5개의 행 출력

train.tail()

Out[15]:

	title	distributor	genre	release_time	time	screening_rat	director	dir_prev_bfnum	dir_prev_num	num_staff	num_actor	box_off_num
595	해무	(주)NEW	드라마	2014-08-13	111	청소년 관람불가	심성보	3833.0	1	510	7	1475091
596	파파로티	(주)쇼박스	드라마	2013-03-14	127	15세 관람가	윤종찬	496061.0	1	286	6	1716438
597	살인의 강	(주)마운틴픽쳐스	공포	2010-09-30	99	청소년 관람불가	김대현	NaN	0	123	4	2475
598	악의 연대기	CJ 엔터테인먼트	느와르	2015-05-14	102	15세 관람가	백운학	NaN	0	431	4	2192525
599	베를린	CJ 엔터테인먼트	액션	2013-01-30	120	15세 관람가	류승완	NaN	0	363	5	7166532

In [16]:

# 데이터 프레임의 행의 개수와 열의 개수 출력

print(train.shape)
print(test.shape)
print(submission.shape)

(600, 12)
(243, 11)
(243, 2)

In [18]:

# train data 정보 요약

train.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 600 entries, 0 to 599
Data columns (total 12 columns):
 #   Column          Non-Null Count  Dtype  
---  ------          --------------  -----  
 0   title           600 non-null    object 
 1   distributor     600 non-null    object 
 2   genre           600 non-null    object 
 3   release_time    600 non-null    object 
 4   time            600 non-null    int64  
 5   screening_rat   600 non-null    object 
 6   director        600 non-null    object 
 7   dir_prev_bfnum  270 non-null    float64
 8   dir_prev_num    600 non-null    int64  
 9   num_staff       600 non-null    int64  
 10  num_actor       600 non-null    int64  
 11  box_off_num     600 non-null    int64  
dtypes: float64(1), int64(5), object(6)
memory usage: 56.4+ KB

In [19]:

# test data 정보 요약

test.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 243 entries, 0 to 242
Data columns (total 11 columns):
 #   Column          Non-Null Count  Dtype  
---  ------          --------------  -----  
 0   title           243 non-null    object 
 1   distributor     243 non-null    object 
 2   genre           243 non-null    object 
 3   release_time    243 non-null    object 
 4   time            243 non-null    int64  
 5   screening_rat   243 non-null    object 
 6   director        243 non-null    object 
 7   dir_prev_bfnum  107 non-null    float64
 8   dir_prev_num    243 non-null    int64  
 9   num_staff       243 non-null    int64  
 10  num_actor       243 non-null    int64  
dtypes: float64(1), int64(4), object(6)
memory usage: 21.0+ KB

In [22]:

# train data 의 퉁계량 출력
# count: 비어있지 않은 값의 개수
# mean: 평균
# std: 표준편차
# min: 최솟값 (이상치(울타리 밖에 있는 부분) 포함)
# 25%: 전체 데이터 정렬했을 때, 하위로부터 1/4번째 지점
# 50%: 전체 데이터 정렬했을 때, 하위로부터 2/4번째 지점
# 75%: 전체 데이터 정렬했을 때, 하위로부터 3/4번째 지점
# max: 최댓값 (이상치 포함)


train.describe()

Out[22]:

	time	dir_prev_bfnum	dir_prev_num	num_staff	num_actor	box_off_num
count	600.000000	2.700000e+02	600.000000	600.000000	600.000000	6.000000e+02
mean	100.863333	1.050443e+06	0.876667	151.118333	3.706667	7.081818e+05
std	18.097528	1.791408e+06	1.183409	165.654671	2.446889	1.828006e+06
min	45.000000	1.000000e+00	0.000000	0.000000	0.000000	1.000000e+00
25%	89.000000	2.038000e+04	0.000000	17.000000	2.000000	1.297250e+03
50%	100.000000	4.784236e+05	0.000000	82.500000	3.000000	1.259100e+04
75%	114.000000	1.286569e+06	2.000000	264.000000	4.000000	4.798868e+05
max	180.000000	1.761531e+07	5.000000	869.000000	25.000000	1.426277e+07

In [23]:

# data 값이 e로 표시되어 읽기 힘드니까
# 소수점 자리수 지정하여 가독성 좋게하는 코드

pd.options.display.float_format = '{:.1f}'.format

In [25]:

# 소수점 지정 포맷 확인

train.describe()

Out[25]:

	time	dir_prev_bfnum	dir_prev_num	num_staff	num_actor	box_off_num
count	600.0	270.0	600.0	600.0	600.0	600.0
mean	100.9	1050442.9	0.9	151.1	3.7	708181.8
std	18.1	1791408.3	1.2	165.7	2.4	1828005.9
min	45.0	1.0	0.0	0.0	0.0	1.0
25%	89.0	20380.0	0.0	17.0	2.0	1297.2
50%	100.0	478423.6	0.0	82.5	3.0	12591.0
75%	114.0	1286568.6	2.0	264.0	4.0	479886.8
max	180.0	17615314.0	5.0	869.0	25.0	14262766.0

In [26]:

# 집단에 대한 통계량 확인
# train data 에서 column = genre, box_off_num 가져오고
# genre를 기준으로 그룹화하여 box_off_num 값 평균내고
# box_off_num 값 기준 오른차순 정렬

train[['genre', 'box_off_num']].groupby('genre').mean().sort_values('box_off_num')

Out[26]:

	box_off_num
genre
뮤지컬	6627.0
다큐멘터리	67172.3
서스펜스	82611.0
애니메이션	181926.7
멜로/로맨스	425968.0
미스터리	527548.2
공포	590832.5
드라마	625689.8
코미디	1193914.0
SF	1788345.7
액션	2203974.1
느와르	2263695.1

In [27]:

# 숫자 소수점 지정하는 포맷팅 리셋

pd.reset_option('display.float_format')

In [28]:

# 리셋된 것 확인

train.describe()

Out[28]:

	time	dir_prev_bfnum	dir_prev_num	num_staff	num_actor	box_off_num
count	600.000000	2.700000e+02	600.000000	600.000000	600.000000	6.000000e+02
mean	100.863333	1.050443e+06	0.876667	151.118333	3.706667	7.081818e+05
std	18.097528	1.791408e+06	1.183409	165.654671	2.446889	1.828006e+06
min	45.000000	1.000000e+00	0.000000	0.000000	0.000000	1.000000e+00
25%	89.000000	2.038000e+04	0.000000	17.000000	2.000000	1.297250e+03
50%	100.000000	4.784236e+05	0.000000	82.500000	3.000000	1.259100e+04
75%	114.000000	1.286569e+06	2.000000	264.000000	4.000000	4.798868e+05
max	180.000000	1.761531e+07	5.000000	869.000000	25.000000	1.426277e+07

In [29]:

# 상관계수(두 개의 변수가 같이 일어나는 강도를 나타내는 수치) 확인
# 보통 0.4 이상이면 두 개의 변수간에 상관성이 있다고 함.

train.corr()

Out[29]:

	time	dir_prev_bfnum	dir_prev_num	num_staff	num_actor	box_off_num
time	1.000000	0.264675	0.306727	0.623205	0.114153	0.441452
dir_prev_bfnum	0.264675	1.000000	0.131822	0.323521	0.083818	0.283184
dir_prev_num	0.306727	0.131822	1.000000	0.450706	0.014006	0.259674
num_staff	0.623205	0.323521	0.450706	1.000000	0.077871	0.544265
num_actor	0.114153	0.083818	0.014006	0.077871	1.000000	0.111179
box_off_num	0.441452	0.283184	0.259674	0.544265	0.111179	1.000000

In [30]:

# 상관성을 heatmap으로 시각화 

import seaborn as sns

sns.heatmap(train.corr(), annot = True)

Out[30]:

<AxesSubplot:>

3. 데이터 전처리¶

Data Cleansing & Pre-Processing¶

In [37]:

# 결측치 확인
# 결측치를 구하고, 행의 수 인 600으로 나누어 줌으로써
# 백분율로 확인

train.isna().sum() / 600

Out[37]:

title             0.00
distributor       0.00
genre             0.00
release_time      0.00
time              0.00
screening_rat     0.00
director          0.00
dir_prev_bfnum    0.55
dir_prev_num      0.00
num_staff         0.00
num_actor         0.00
box_off_num       0.00
dtype: float64

In [39]:

# train data의 dir_prev_bfnum 에서 결측치 있는 행들만 가져옴
# dir_prev_num 값 총합 출력

train[train['dir_prev_bfnum'].isna()]['dir_prev_num'].sum()

Out[39]:

In [41]:

# 기존 결측치 값 Nan 으로 표기 되어 있음
# Nan 을 0 으로 바꿔주는 코드

train['dir_prev_bfnum'].fillna(0, inplace = True)

In [46]:

# 결측치 값 변환 된것 확인

train.isna().sum()

Out[46]:

title             0
distributor       0
genre             0
release_time      0
time              0
screening_rat     0
director          0
dir_prev_bfnum    0
dir_prev_num      0
num_staff         0
num_actor         0
box_off_num       0
dtype: int64

In [47]:

# test 결측치 값도 동일하게 위의 과정으로 변환

test[test['dir_prev_bfnum'].isna()]['dir_prev_num'].sum()

Out[47]:

In [48]:

# test 결측치 값도 동일하게 위의 과정으로 변환

test['dir_prev_bfnum'].fillna(0, inplace = True)

4. 변수 선택 및 모델 구축¶

Feature Engineering & Initial Modeling¶

In [49]:

# model 설계

model = lgb.LGBMRegressor(random_state = 777, n_estimators = 1000)

In [50]:

# features 는 x, target 은 y

features = ['time', 'dir_prev_num', 'num_staff', 'num_actor']
target = ['box_off_num']

In [51]:

# train data와 test data 변수에 입력

X_train, X_test, y_train = train[features], test[features], train[target]

5. 모델 학습 및 검증¶

Model Tuning & Evaluation¶

a. lightGBM (base model)¶

b. k-fold lightGBM (k-fold model)¶

c. feature engineering (fe)¶

d. grid search (hyperparameter tuning)¶

LightGBM¶

In [52]:

# train data 학습 

model.fit(X_train, y_train)

Out[52]:

LGBMRegressor(n_estimators=1000, random_state=777)

In [53]:

# submission data 사본 변수에 입력

singleLGBM = submission.copy()

In [55]:

# singleLGBM data 상위 5개의 행 출력

singleLGBM.head()

Out[55]:

	title	box_off_num
0	용서는 없다	0
1	아빠가 여자를 좋아해	0
2	하모니	0
3	의형제	0
4	평행 이론	0

In [61]:

# 예측한 결과값을 singleLGBM['box_off_num'] 에 입력
singleLGBM['box_off_num'] = model.predict(X_test)

# 가독성을 위해 숫자 소수점 지정
pd.options.display.float_format = '{:.1f}'.format

# 확인
singleLGBM

Out[61]:

	title	box_off_num
0	용서는 없다	2817995.2
1	아빠가 여자를 좋아해	375377.2
2	하모니	-569324.3
3	의형제	1581189.0
4	평행 이론	-527780.6
...	...	...
238	해에게서 소년에게	500784.4
239	울보 권투부	1013858.4
240	어떤살인	1682067.7
241	말하지 못한 비밀	300216.3
242	조선안방 스캔들-칠거지악 2	11390.0

243 rows × 2 columns

In [62]:

# csv 파일 만들어서, 결과물 저장

singleLGBM.to_csv('data/singleLGBM.csv', index = False)

k-fold 교차 검증 (k-fold cross validation)¶

In [66]:

# 패키지 불러오기
from sklearn.model_selection import KFold

# 데이터 분할 수 = 5, 데이터 shuffle 하기
k_fold = KFold(n_splits = 5, shuffle = True, random_state = 777)

In [67]:

# X_train data 에서 train 480개, val 120개로 분할

for train_idx, val_idx in k_fold.split(X_train):
    print(len(train_idx), len(val_idx))
    break

480 120

In [68]:

# 교차 검증하여, LGBMRegressor 모델 학습시키기

model = lgb.LGBMRegressor(random_state = 777, n_estimators = 1000)

models = []

for train_idx, val_idx in k_fold.split(X_train):
    x_t = X_train.iloc[train_idx]
    y_t = y_train.iloc[train_idx]
    x_val = X_train.iloc[val_idx]
    y_val = y_train.iloc[val_idx]
    
    models.append(model.fit(x_t, y_t, eval_set=(x_val, y_val), early_stopping_rounds = 100, verbose = 100))

C:\Users\Jung_dayoung\anaconda3\lib\site-packages\lightgbm\sklearn.py:726: UserWarning: 'early_stopping_rounds' argument is deprecated and will be removed in a future release of LightGBM. Pass 'early_stopping()' callback via 'callbacks' argument instead.
  _log_warning("'early_stopping_rounds' argument is deprecated and will be removed in a future release of LightGBM. "
C:\Users\Jung_dayoung\anaconda3\lib\site-packages\lightgbm\sklearn.py:736: UserWarning: 'verbose' argument is deprecated and will be removed in a future release of LightGBM. Pass 'log_evaluation()' callback via 'callbacks' argument instead.
  _log_warning("'verbose' argument is deprecated and will be removed in a future release of LightGBM. "

[100]	valid_0's l2: 2.70572e+12
[100]	valid_0's l2: 3.90847e+12
[100]	valid_0's l2: 3.50344e+12
[100]	valid_0's l2: 1.45977e+12
[100]	valid_0's l2: 1.77214e+12

In [69]:

models

Out[69]:

[LGBMRegressor(n_estimators=1000, random_state=777),
 LGBMRegressor(n_estimators=1000, random_state=777),
 LGBMRegressor(n_estimators=1000, random_state=777),
 LGBMRegressor(n_estimators=1000, random_state=777),
 LGBMRegressor(n_estimators=1000, random_state=777)]

In [71]:

# 예측 결과값 정리

preds = []

for model in models:
    preds.append(model.predict(X_test))
len(preds)

Out[71]:

In [72]:

# 예측 결과값 확인

preds

Out[72]:

[array([3367422.08211024,  961138.88337016, 1097929.67851313,
        2097270.81999921,  781476.31131047,  123133.23396977,
          84085.01896248,  199222.92670303,  124854.97973097,
        1072684.78820647,  657040.55769984, 1644701.47160779,
         924735.72080619,  191699.58415836,  723218.39948755,
        2395613.42888462,  113180.54906592, 1882765.92812296,
          97286.47282983,  417270.42113431,  161482.3367577 ,
          63495.73363115,  598057.42337284,  249411.25868881,
         308424.71884273, 1319518.60852241,  202273.52740684,
        1076683.13768137,  491636.41945325,  183406.6380314 ,
        1740233.45816734,   86696.17528125,  440430.29773088,
        2040944.13061099,  113375.45695331,  531746.29484356,
         106207.35971699,  169560.79581187,  158101.0895316 ,
          95037.29701059,  466355.50624629,  197548.63265325,
         198810.0901353 ,  682743.4340129 ,  679265.22397749,
         617812.91125278,  219182.85833837,   83712.10602149,
        1206257.86593038,  213886.45118276,  225829.13565632,
         573638.25784044,  213008.72404464, 1256363.95636108,
        1114803.4971861 ,  743531.85320002,  491360.56067236,
         844903.11979141, 1294842.58551457, 3224861.58056056,
         620636.85839868, 1997931.01104478,  425408.70724227,
         187911.96484382,  146932.14876274,   60290.34778046,
        1553051.98103674,   61273.32234969,  434403.58150991,
         295641.77082639, 3270724.75199692,  956695.86295249,
          52421.39380137, 1193829.08957371,  109115.94774032,
         114532.36574566, 1889840.31759081,  263117.7221664 ,
         110722.44800397,  129405.69463079,  295288.27657346,
        2208289.14524844,  127156.40131508, 2356552.01519359,
          52421.39380137, 1582094.56763018,  450263.02347743,
         180314.7209139 ,  207866.80825891,  237980.96274482,
          92809.10191796,  217726.13148671,  588167.04361702,
        2269569.15414991,  413549.08320869,  504238.61127942,
         121235.38111888,  294348.6064813 ,  365600.03475309,
        1021310.81474734, 1255348.78654001, 1121695.37239584,
         236858.1355381 ,  180837.57476421,  390892.54870688,
         112045.06953824, 3091294.61560971,   71489.84756224,
        3949551.73068699,  117269.53815136,   76582.3994148 ,
        1933601.27793109, 1675206.65522009,  294348.6064813 ,
          97940.27258456,  945839.01956416,  610770.20862757,
        1804885.34137216,  199766.90646407,   70766.27122776,
          81754.00329429, 1021310.81474734,  846364.16218315,
          77206.4852859 ,  598474.09814662,  169200.18001164,
          73757.25966134,   68560.19329018,  363573.18970532,
         439100.02360785,   53273.19743239,  883179.93430519,
          96841.00528099,   69705.57216142,  720189.94864194,
        4682200.86959811,  157547.86592702,  598577.33917498,
         103530.67260964,  219913.31885361,   69705.57216142,
        3393838.48355355, 3747750.51209052,  137237.96352743,
         205694.9146793 , 1343063.74316061,  105778.55398733,
         113406.67661411, 2090467.44678197,  676486.18009377,
        1653182.84275506,  151567.24115244,   82132.69602781,
         450200.37816199,  123824.9061514 ,  841869.56483077,
          95800.41312684,  118291.73959934, 2085585.62459758,
        2372839.45720666,  247688.59858277, 1334681.62533764,
         101692.7552419 ,  503472.70498147,   64636.19839218,
         320805.50593895, 2623266.33617439,  174088.14927962,
          72116.10600518,   80293.56098003,   89828.86338213,
         219193.02374581,  328023.58124177,   60290.34778046,
         516294.37658866,  236774.09865655,  163101.99672652,
          95037.29701059, 4401076.73195112, 3540272.0939466 ,
          80652.10214941,   64007.05829165,   91088.74649744,
         451288.50611234, 2108891.7577948 ,  222324.39435632,
         230969.76551419,  119869.73646973,  202629.14049356,
        3372525.98485469,  194055.24146137,   82570.78055486,
         107755.73232124,   98100.67856341,   64007.05829165,
        1960017.88124655,  439387.54981251, 1311678.25959343,
          52266.56751161,  612357.4359533 ,  277611.14005727,
         100834.87848026,   52421.39380137,  320518.12600116,
         622991.23928914,  905631.603816  , 2927834.24849227,
         562358.56357488,  111381.83047422,  781476.31131047,
         798596.64645281, 2310966.59797285, 5133725.87086224,
        1281440.30372974,  171353.72979582,   89630.73883889,
        1555941.58912764,  124073.12498191,   66108.35792552,
         201459.85880026,  108993.88378627, 3134539.90969621,
         148239.45542642,   53451.2389687 , 1307718.56304709,
         214406.20173983,  194835.86837801, 3367422.08211024,
         199766.90646407, 3823577.41600747,   68447.35094925,
         739960.30606392,  506712.38290343,  150255.49362263,
          62093.65018694,   64707.81660264,  151784.52074685,
          80757.16568265,  177095.84545845,  210255.47503947,
         360547.62990175,  181508.22558258,   55171.21699634]),
 array([3367422.08211024,  961138.88337016, 1097929.67851313,
        2097270.81999921,  781476.31131047,  123133.23396977,
          84085.01896248,  199222.92670303,  124854.97973097,
        1072684.78820647,  657040.55769984, 1644701.47160779,
         924735.72080619,  191699.58415836,  723218.39948755,
        2395613.42888462,  113180.54906592, 1882765.92812296,
          97286.47282983,  417270.42113431,  161482.3367577 ,
          63495.73363115,  598057.42337284,  249411.25868881,
         308424.71884273, 1319518.60852241,  202273.52740684,
        1076683.13768137,  491636.41945325,  183406.6380314 ,
        1740233.45816734,   86696.17528125,  440430.29773088,
        2040944.13061099,  113375.45695331,  531746.29484356,
         106207.35971699,  169560.79581187,  158101.0895316 ,
          95037.29701059,  466355.50624629,  197548.63265325,
         198810.0901353 ,  682743.4340129 ,  679265.22397749,
         617812.91125278,  219182.85833837,   83712.10602149,
        1206257.86593038,  213886.45118276,  225829.13565632,
         573638.25784044,  213008.72404464, 1256363.95636108,
        1114803.4971861 ,  743531.85320002,  491360.56067236,
         844903.11979141, 1294842.58551457, 3224861.58056056,
         620636.85839868, 1997931.01104478,  425408.70724227,
         187911.96484382,  146932.14876274,   60290.34778046,
        1553051.98103674,   61273.32234969,  434403.58150991,
         295641.77082639, 3270724.75199692,  956695.86295249,
          52421.39380137, 1193829.08957371,  109115.94774032,
         114532.36574566, 1889840.31759081,  263117.7221664 ,
         110722.44800397,  129405.69463079,  295288.27657346,
        2208289.14524844,  127156.40131508, 2356552.01519359,
          52421.39380137, 1582094.56763018,  450263.02347743,
         180314.7209139 ,  207866.80825891,  237980.96274482,
          92809.10191796,  217726.13148671,  588167.04361702,
        2269569.15414991,  413549.08320869,  504238.61127942,
         121235.38111888,  294348.6064813 ,  365600.03475309,
        1021310.81474734, 1255348.78654001, 1121695.37239584,
         236858.1355381 ,  180837.57476421,  390892.54870688,
         112045.06953824, 3091294.61560971,   71489.84756224,
        3949551.73068699,  117269.53815136,   76582.3994148 ,
        1933601.27793109, 1675206.65522009,  294348.6064813 ,
          97940.27258456,  945839.01956416,  610770.20862757,
        1804885.34137216,  199766.90646407,   70766.27122776,
          81754.00329429, 1021310.81474734,  846364.16218315,
          77206.4852859 ,  598474.09814662,  169200.18001164,
          73757.25966134,   68560.19329018,  363573.18970532,
         439100.02360785,   53273.19743239,  883179.93430519,
          96841.00528099,   69705.57216142,  720189.94864194,
        4682200.86959811,  157547.86592702,  598577.33917498,
         103530.67260964,  219913.31885361,   69705.57216142,
        3393838.48355355, 3747750.51209052,  137237.96352743,
         205694.9146793 , 1343063.74316061,  105778.55398733,
         113406.67661411, 2090467.44678197,  676486.18009377,
        1653182.84275506,  151567.24115244,   82132.69602781,
         450200.37816199,  123824.9061514 ,  841869.56483077,
          95800.41312684,  118291.73959934, 2085585.62459758,
        2372839.45720666,  247688.59858277, 1334681.62533764,
         101692.7552419 ,  503472.70498147,   64636.19839218,
         320805.50593895, 2623266.33617439,  174088.14927962,
          72116.10600518,   80293.56098003,   89828.86338213,
         219193.02374581,  328023.58124177,   60290.34778046,
         516294.37658866,  236774.09865655,  163101.99672652,
          95037.29701059, 4401076.73195112, 3540272.0939466 ,
          80652.10214941,   64007.05829165,   91088.74649744,
         451288.50611234, 2108891.7577948 ,  222324.39435632,
         230969.76551419,  119869.73646973,  202629.14049356,
        3372525.98485469,  194055.24146137,   82570.78055486,
         107755.73232124,   98100.67856341,   64007.05829165,
        1960017.88124655,  439387.54981251, 1311678.25959343,
          52266.56751161,  612357.4359533 ,  277611.14005727,
         100834.87848026,   52421.39380137,  320518.12600116,
         622991.23928914,  905631.603816  , 2927834.24849227,
         562358.56357488,  111381.83047422,  781476.31131047,
         798596.64645281, 2310966.59797285, 5133725.87086224,
        1281440.30372974,  171353.72979582,   89630.73883889,
        1555941.58912764,  124073.12498191,   66108.35792552,
         201459.85880026,  108993.88378627, 3134539.90969621,
         148239.45542642,   53451.2389687 , 1307718.56304709,
         214406.20173983,  194835.86837801, 3367422.08211024,
         199766.90646407, 3823577.41600747,   68447.35094925,
         739960.30606392,  506712.38290343,  150255.49362263,
          62093.65018694,   64707.81660264,  151784.52074685,
          80757.16568265,  177095.84545845,  210255.47503947,
         360547.62990175,  181508.22558258,   55171.21699634]),
 array([3367422.08211024,  961138.88337016, 1097929.67851313,
        2097270.81999921,  781476.31131047,  123133.23396977,
          84085.01896248,  199222.92670303,  124854.97973097,
        1072684.78820647,  657040.55769984, 1644701.47160779,
         924735.72080619,  191699.58415836,  723218.39948755,
        2395613.42888462,  113180.54906592, 1882765.92812296,
          97286.47282983,  417270.42113431,  161482.3367577 ,
          63495.73363115,  598057.42337284,  249411.25868881,
         308424.71884273, 1319518.60852241,  202273.52740684,
        1076683.13768137,  491636.41945325,  183406.6380314 ,
        1740233.45816734,   86696.17528125,  440430.29773088,
        2040944.13061099,  113375.45695331,  531746.29484356,
         106207.35971699,  169560.79581187,  158101.0895316 ,
          95037.29701059,  466355.50624629,  197548.63265325,
         198810.0901353 ,  682743.4340129 ,  679265.22397749,
         617812.91125278,  219182.85833837,   83712.10602149,
        1206257.86593038,  213886.45118276,  225829.13565632,
         573638.25784044,  213008.72404464, 1256363.95636108,
        1114803.4971861 ,  743531.85320002,  491360.56067236,
         844903.11979141, 1294842.58551457, 3224861.58056056,
         620636.85839868, 1997931.01104478,  425408.70724227,
         187911.96484382,  146932.14876274,   60290.34778046,
        1553051.98103674,   61273.32234969,  434403.58150991,
         295641.77082639, 3270724.75199692,  956695.86295249,
          52421.39380137, 1193829.08957371,  109115.94774032,
         114532.36574566, 1889840.31759081,  263117.7221664 ,
         110722.44800397,  129405.69463079,  295288.27657346,
        2208289.14524844,  127156.40131508, 2356552.01519359,
          52421.39380137, 1582094.56763018,  450263.02347743,
         180314.7209139 ,  207866.80825891,  237980.96274482,
          92809.10191796,  217726.13148671,  588167.04361702,
        2269569.15414991,  413549.08320869,  504238.61127942,
         121235.38111888,  294348.6064813 ,  365600.03475309,
        1021310.81474734, 1255348.78654001, 1121695.37239584,
         236858.1355381 ,  180837.57476421,  390892.54870688,
         112045.06953824, 3091294.61560971,   71489.84756224,
        3949551.73068699,  117269.53815136,   76582.3994148 ,
        1933601.27793109, 1675206.65522009,  294348.6064813 ,
          97940.27258456,  945839.01956416,  610770.20862757,
        1804885.34137216,  199766.90646407,   70766.27122776,
          81754.00329429, 1021310.81474734,  846364.16218315,
          77206.4852859 ,  598474.09814662,  169200.18001164,
          73757.25966134,   68560.19329018,  363573.18970532,
         439100.02360785,   53273.19743239,  883179.93430519,
          96841.00528099,   69705.57216142,  720189.94864194,
        4682200.86959811,  157547.86592702,  598577.33917498,
         103530.67260964,  219913.31885361,   69705.57216142,
        3393838.48355355, 3747750.51209052,  137237.96352743,
         205694.9146793 , 1343063.74316061,  105778.55398733,
         113406.67661411, 2090467.44678197,  676486.18009377,
        1653182.84275506,  151567.24115244,   82132.69602781,
         450200.37816199,  123824.9061514 ,  841869.56483077,
          95800.41312684,  118291.73959934, 2085585.62459758,
        2372839.45720666,  247688.59858277, 1334681.62533764,
         101692.7552419 ,  503472.70498147,   64636.19839218,
         320805.50593895, 2623266.33617439,  174088.14927962,
          72116.10600518,   80293.56098003,   89828.86338213,
         219193.02374581,  328023.58124177,   60290.34778046,
         516294.37658866,  236774.09865655,  163101.99672652,
          95037.29701059, 4401076.73195112, 3540272.0939466 ,
          80652.10214941,   64007.05829165,   91088.74649744,
         451288.50611234, 2108891.7577948 ,  222324.39435632,
         230969.76551419,  119869.73646973,  202629.14049356,
        3372525.98485469,  194055.24146137,   82570.78055486,
         107755.73232124,   98100.67856341,   64007.05829165,
        1960017.88124655,  439387.54981251, 1311678.25959343,
          52266.56751161,  612357.4359533 ,  277611.14005727,
         100834.87848026,   52421.39380137,  320518.12600116,
         622991.23928914,  905631.603816  , 2927834.24849227,
         562358.56357488,  111381.83047422,  781476.31131047,
         798596.64645281, 2310966.59797285, 5133725.87086224,
        1281440.30372974,  171353.72979582,   89630.73883889,
        1555941.58912764,  124073.12498191,   66108.35792552,
         201459.85880026,  108993.88378627, 3134539.90969621,
         148239.45542642,   53451.2389687 , 1307718.56304709,
         214406.20173983,  194835.86837801, 3367422.08211024,
         199766.90646407, 3823577.41600747,   68447.35094925,
         739960.30606392,  506712.38290343,  150255.49362263,
          62093.65018694,   64707.81660264,  151784.52074685,
          80757.16568265,  177095.84545845,  210255.47503947,
         360547.62990175,  181508.22558258,   55171.21699634]),
 array([3367422.08211024,  961138.88337016, 1097929.67851313,
        2097270.81999921,  781476.31131047,  123133.23396977,
          84085.01896248,  199222.92670303,  124854.97973097,
        1072684.78820647,  657040.55769984, 1644701.47160779,
         924735.72080619,  191699.58415836,  723218.39948755,
        2395613.42888462,  113180.54906592, 1882765.92812296,
          97286.47282983,  417270.42113431,  161482.3367577 ,
          63495.73363115,  598057.42337284,  249411.25868881,
         308424.71884273, 1319518.60852241,  202273.52740684,
        1076683.13768137,  491636.41945325,  183406.6380314 ,
        1740233.45816734,   86696.17528125,  440430.29773088,
        2040944.13061099,  113375.45695331,  531746.29484356,
         106207.35971699,  169560.79581187,  158101.0895316 ,
          95037.29701059,  466355.50624629,  197548.63265325,
         198810.0901353 ,  682743.4340129 ,  679265.22397749,
         617812.91125278,  219182.85833837,   83712.10602149,
        1206257.86593038,  213886.45118276,  225829.13565632,
         573638.25784044,  213008.72404464, 1256363.95636108,
        1114803.4971861 ,  743531.85320002,  491360.56067236,
         844903.11979141, 1294842.58551457, 3224861.58056056,
         620636.85839868, 1997931.01104478,  425408.70724227,
         187911.96484382,  146932.14876274,   60290.34778046,
        1553051.98103674,   61273.32234969,  434403.58150991,
         295641.77082639, 3270724.75199692,  956695.86295249,
          52421.39380137, 1193829.08957371,  109115.94774032,
         114532.36574566, 1889840.31759081,  263117.7221664 ,
         110722.44800397,  129405.69463079,  295288.27657346,
        2208289.14524844,  127156.40131508, 2356552.01519359,
          52421.39380137, 1582094.56763018,  450263.02347743,
         180314.7209139 ,  207866.80825891,  237980.96274482,
          92809.10191796,  217726.13148671,  588167.04361702,
        2269569.15414991,  413549.08320869,  504238.61127942,
         121235.38111888,  294348.6064813 ,  365600.03475309,
        1021310.81474734, 1255348.78654001, 1121695.37239584,
         236858.1355381 ,  180837.57476421,  390892.54870688,
         112045.06953824, 3091294.61560971,   71489.84756224,
        3949551.73068699,  117269.53815136,   76582.3994148 ,
        1933601.27793109, 1675206.65522009,  294348.6064813 ,
          97940.27258456,  945839.01956416,  610770.20862757,
        1804885.34137216,  199766.90646407,   70766.27122776,
          81754.00329429, 1021310.81474734,  846364.16218315,
          77206.4852859 ,  598474.09814662,  169200.18001164,
          73757.25966134,   68560.19329018,  363573.18970532,
         439100.02360785,   53273.19743239,  883179.93430519,
          96841.00528099,   69705.57216142,  720189.94864194,
        4682200.86959811,  157547.86592702,  598577.33917498,
         103530.67260964,  219913.31885361,   69705.57216142,
        3393838.48355355, 3747750.51209052,  137237.96352743,
         205694.9146793 , 1343063.74316061,  105778.55398733,
         113406.67661411, 2090467.44678197,  676486.18009377,
        1653182.84275506,  151567.24115244,   82132.69602781,
         450200.37816199,  123824.9061514 ,  841869.56483077,
          95800.41312684,  118291.73959934, 2085585.62459758,
        2372839.45720666,  247688.59858277, 1334681.62533764,
         101692.7552419 ,  503472.70498147,   64636.19839218,
         320805.50593895, 2623266.33617439,  174088.14927962,
          72116.10600518,   80293.56098003,   89828.86338213,
         219193.02374581,  328023.58124177,   60290.34778046,
         516294.37658866,  236774.09865655,  163101.99672652,
          95037.29701059, 4401076.73195112, 3540272.0939466 ,
          80652.10214941,   64007.05829165,   91088.74649744,
         451288.50611234, 2108891.7577948 ,  222324.39435632,
         230969.76551419,  119869.73646973,  202629.14049356,
        3372525.98485469,  194055.24146137,   82570.78055486,
         107755.73232124,   98100.67856341,   64007.05829165,
        1960017.88124655,  439387.54981251, 1311678.25959343,
          52266.56751161,  612357.4359533 ,  277611.14005727,
         100834.87848026,   52421.39380137,  320518.12600116,
         622991.23928914,  905631.603816  , 2927834.24849227,
         562358.56357488,  111381.83047422,  781476.31131047,
         798596.64645281, 2310966.59797285, 5133725.87086224,
        1281440.30372974,  171353.72979582,   89630.73883889,
        1555941.58912764,  124073.12498191,   66108.35792552,
         201459.85880026,  108993.88378627, 3134539.90969621,
         148239.45542642,   53451.2389687 , 1307718.56304709,
         214406.20173983,  194835.86837801, 3367422.08211024,
         199766.90646407, 3823577.41600747,   68447.35094925,
         739960.30606392,  506712.38290343,  150255.49362263,
          62093.65018694,   64707.81660264,  151784.52074685,
          80757.16568265,  177095.84545845,  210255.47503947,
         360547.62990175,  181508.22558258,   55171.21699634]),
 array([3367422.08211024,  961138.88337016, 1097929.67851313,
        2097270.81999921,  781476.31131047,  123133.23396977,
          84085.01896248,  199222.92670303,  124854.97973097,
        1072684.78820647,  657040.55769984, 1644701.47160779,
         924735.72080619,  191699.58415836,  723218.39948755,
        2395613.42888462,  113180.54906592, 1882765.92812296,
          97286.47282983,  417270.42113431,  161482.3367577 ,
          63495.73363115,  598057.42337284,  249411.25868881,
         308424.71884273, 1319518.60852241,  202273.52740684,
        1076683.13768137,  491636.41945325,  183406.6380314 ,
        1740233.45816734,   86696.17528125,  440430.29773088,
        2040944.13061099,  113375.45695331,  531746.29484356,
         106207.35971699,  169560.79581187,  158101.0895316 ,
          95037.29701059,  466355.50624629,  197548.63265325,
         198810.0901353 ,  682743.4340129 ,  679265.22397749,
         617812.91125278,  219182.85833837,   83712.10602149,
        1206257.86593038,  213886.45118276,  225829.13565632,
         573638.25784044,  213008.72404464, 1256363.95636108,
        1114803.4971861 ,  743531.85320002,  491360.56067236,
         844903.11979141, 1294842.58551457, 3224861.58056056,
         620636.85839868, 1997931.01104478,  425408.70724227,
         187911.96484382,  146932.14876274,   60290.34778046,
        1553051.98103674,   61273.32234969,  434403.58150991,
         295641.77082639, 3270724.75199692,  956695.86295249,
          52421.39380137, 1193829.08957371,  109115.94774032,
         114532.36574566, 1889840.31759081,  263117.7221664 ,
         110722.44800397,  129405.69463079,  295288.27657346,
        2208289.14524844,  127156.40131508, 2356552.01519359,
          52421.39380137, 1582094.56763018,  450263.02347743,
         180314.7209139 ,  207866.80825891,  237980.96274482,
          92809.10191796,  217726.13148671,  588167.04361702,
        2269569.15414991,  413549.08320869,  504238.61127942,
         121235.38111888,  294348.6064813 ,  365600.03475309,
        1021310.81474734, 1255348.78654001, 1121695.37239584,
         236858.1355381 ,  180837.57476421,  390892.54870688,
         112045.06953824, 3091294.61560971,   71489.84756224,
        3949551.73068699,  117269.53815136,   76582.3994148 ,
        1933601.27793109, 1675206.65522009,  294348.6064813 ,
          97940.27258456,  945839.01956416,  610770.20862757,
        1804885.34137216,  199766.90646407,   70766.27122776,
          81754.00329429, 1021310.81474734,  846364.16218315,
          77206.4852859 ,  598474.09814662,  169200.18001164,
          73757.25966134,   68560.19329018,  363573.18970532,
         439100.02360785,   53273.19743239,  883179.93430519,
          96841.00528099,   69705.57216142,  720189.94864194,
        4682200.86959811,  157547.86592702,  598577.33917498,
         103530.67260964,  219913.31885361,   69705.57216142,
        3393838.48355355, 3747750.51209052,  137237.96352743,
         205694.9146793 , 1343063.74316061,  105778.55398733,
         113406.67661411, 2090467.44678197,  676486.18009377,
        1653182.84275506,  151567.24115244,   82132.69602781,
         450200.37816199,  123824.9061514 ,  841869.56483077,
          95800.41312684,  118291.73959934, 2085585.62459758,
        2372839.45720666,  247688.59858277, 1334681.62533764,
         101692.7552419 ,  503472.70498147,   64636.19839218,
         320805.50593895, 2623266.33617439,  174088.14927962,
          72116.10600518,   80293.56098003,   89828.86338213,
         219193.02374581,  328023.58124177,   60290.34778046,
         516294.37658866,  236774.09865655,  163101.99672652,
          95037.29701059, 4401076.73195112, 3540272.0939466 ,
          80652.10214941,   64007.05829165,   91088.74649744,
         451288.50611234, 2108891.7577948 ,  222324.39435632,
         230969.76551419,  119869.73646973,  202629.14049356,
        3372525.98485469,  194055.24146137,   82570.78055486,
         107755.73232124,   98100.67856341,   64007.05829165,
        1960017.88124655,  439387.54981251, 1311678.25959343,
          52266.56751161,  612357.4359533 ,  277611.14005727,
         100834.87848026,   52421.39380137,  320518.12600116,
         622991.23928914,  905631.603816  , 2927834.24849227,
         562358.56357488,  111381.83047422,  781476.31131047,
         798596.64645281, 2310966.59797285, 5133725.87086224,
        1281440.30372974,  171353.72979582,   89630.73883889,
        1555941.58912764,  124073.12498191,   66108.35792552,
         201459.85880026,  108993.88378627, 3134539.90969621,
         148239.45542642,   53451.2389687 , 1307718.56304709,
         214406.20173983,  194835.86837801, 3367422.08211024,
         199766.90646407, 3823577.41600747,   68447.35094925,
         739960.30606392,  506712.38290343,  150255.49362263,
          62093.65018694,   64707.81660264,  151784.52074685,
          80757.16568265,  177095.84545845,  210255.47503947,
         360547.62990175,  181508.22558258,   55171.21699634])]

In [73]:

# submission 사본을 kfoldLightGBM 입력
kfoldLightGBM = submission.copy()


# 예측 결과값 kfoldLightGBM['box_off_num'] 에 추가
import numpy as np
kfoldLightGBM['box_off_num'] = np.mean(preds, axis = 0)


# kfoldLightGBM 상위 5개의 행 출력
kfoldLightGBM.head()

Out[73]:

	title	box_off_num
0	용서는 없다	3367422.1
1	아빠가 여자를 좋아해	961138.9
2	하모니	1097929.7
3	의형제	2097270.8
4	평행 이론	781476.3

In [76]:

# kfoldLightGBM.csv 파일 만든 후 결과값 저장

kfoldLightGBM.to_csv('data/kfoldLightGBM.csv', index = False)

feature engineering¶

In [77]:

features

Out[77]:

['time', 'dir_prev_num', 'num_staff', 'num_actor']

In [80]:

train.columns

Out[80]:

Index(['title', 'distributor', 'genre', 'release_time', 'time',
       'screening_rat', 'director', 'dir_prev_bfnum', 'dir_prev_num',
       'num_staff', 'num_actor', 'box_off_num'],
      dtype='object')

In [81]:

train.genre

Out[81]:

0       액션
1      느와르
2       액션
3      코미디
4      코미디
      ... 
595    드라마
596    드라마
597     공포
598    느와르
599     액션
Name: genre, Length: 600, dtype: object

In [82]:

# train data의 genre 값을 정수로 변환

from sklearn import preprocessing
le = preprocessing.LabelEncoder()
train['genre'] = le.fit_transform(train['genre'])

train['genre']

Out[82]:

0      10
1       2
2      10
3      11
4      11
       ..
595     4
596     4
597     1
598     2
599    10
Name: genre, Length: 600, dtype: int32

In [90]:

# test data의 genre 값을 정수로 변환

test['genre'] = le.transform(test['genre'])
test['genre']

Out[90]:

0       2
1       5
2       4
3      10
4       1
       ..
238     4
239     3
240     2
241     4
242     5
Name: genre, Length: 243, dtype: int32

In [94]:

# features 값 재정의

features = ['time', 'dir_prev_num', 'num_staff', 'num_actor', 'dir_prev_bfnum', 'genre']

In [95]:

# 학습을 위해 필요한 변수에 값 할당

X_train, X_test, y_train = train[features], test[features], train[target]

In [96]:

# k-fold 와 LGBMRegressor 로 학습 시키기

model = lgb.LGBMRegressor(random_state = 777, n_estimators = 1000)

models = []

for train_idx, val_idx in k_fold.split(X_train):
    x_t = X_train.iloc[train_idx]
    y_t = y_train.iloc[train_idx]
    x_val = X_train.iloc[val_idx]
    y_val = y_train.iloc[val_idx]
    
    models.append(model.fit(x_t, y_t, eval_set = (x_val, y_val), early_stopping_rounds = 100, verbose = 100))

C:\Users\Jung_dayoung\anaconda3\lib\site-packages\lightgbm\sklearn.py:726: UserWarning: 'early_stopping_rounds' argument is deprecated and will be removed in a future release of LightGBM. Pass 'early_stopping()' callback via 'callbacks' argument instead.
  _log_warning("'early_stopping_rounds' argument is deprecated and will be removed in a future release of LightGBM. "
C:\Users\Jung_dayoung\anaconda3\lib\site-packages\lightgbm\sklearn.py:736: UserWarning: 'verbose' argument is deprecated and will be removed in a future release of LightGBM. Pass 'log_evaluation()' callback via 'callbacks' argument instead.
  _log_warning("'verbose' argument is deprecated and will be removed in a future release of LightGBM. "

[100]	valid_0's l2: 2.62067e+12
[100]	valid_0's l2: 4.39227e+12
[100]	valid_0's l2: 3.29841e+12
[100]	valid_0's l2: 1.56499e+12
[100]	valid_0's l2: 1.60118e+12

In [97]:

X_test.head()

Out[97]:

	time	dir_prev_num	num_staff	num_actor	dir_prev_bfnum	genre
0	125	2	304	3	300529.0	2
1	113	4	275	3	342700.2	5
2	115	3	419	7	4206610.7	4
3	116	2	408	2	691342.0	10
4	110	1	380	1	31738.0	1

In [98]:

# X_test로 결과값 예측

preds = []

for model in models:
    preds.append(model.predict(X_test))
len(preds)

Out[98]:

In [101]:

# submission 사본을 feLightGBM 에 할당
feLightGBM = submission.copy()


# 예측값의 평균을 feLightGBM['box_off_num'] 에 할당
feLightGBM['box_off_num'] = np.mean(preds, axis = 0)


# feLightGBM.csv 파일 만든 후 결과값 저장
feLightGBM.to_csv('data/feLightGBM.csv', index = False)

Grid Search¶

가장 우수한 성능을 보이는 모델의 하이퍼 파라미터를 찾기 위해서¶

In [102]:

# 필요한 패키지 불러오기
from sklearn.model_selection import GridSearchCV

In [105]:

# 최적화된 파라미터 찾기

model = lgb.LGBMRegressor(random_state = 777, n_estimators = 1000)

params = {
    'learning_rate': [0.1, 0.01, 0.003],
    'min_child_samples': [20, 30]    
}

gs = GridSearchCV(estimator = model,
                 param_grid = params,
                 scoring = 'neg_mean_squared_error',
                 cv = k_fold)

In [106]:

# 학습시키기
gs.fit(X_train, y_train)

Out[106]:

GridSearchCV(cv=KFold(n_splits=5, random_state=777, shuffle=True),
             estimator=LGBMRegressor(n_estimators=1000, random_state=777),
             param_grid={'learning_rate': [0.1, 0.01, 0.003],
                         'min_child_samples': [20, 30]},
             scoring='neg_mean_squared_error')

In [107]:

# 최적화된 파라미터 출력
gs.best_params_

Out[107]:

{'learning_rate': 0.003, 'min_child_samples': 30}

In [108]:

# 학습 시키기

model = lgb.LGBMRegressor(random_state = 777, n_estimators = 1000, 
                          learning_rate = 0.003, min_child_samples = 30)

models = []

for train_idx, val_idx in k_fold.split(X_train):
    x_t = X_train.iloc[train_idx]
    y_t = y_train.iloc[train_idx]
    x_val = X_train.iloc[val_idx]
    y_val = y_train.iloc[val_idx]
    
    models.append(model.fit(x_t, y_t, eval_set = (x_val, y_val), early_stopping_rounds = 100, verbose = 100))

C:\Users\Jung_dayoung\anaconda3\lib\site-packages\lightgbm\sklearn.py:726: UserWarning: 'early_stopping_rounds' argument is deprecated and will be removed in a future release of LightGBM. Pass 'early_stopping()' callback via 'callbacks' argument instead.
  _log_warning("'early_stopping_rounds' argument is deprecated and will be removed in a future release of LightGBM. "
C:\Users\Jung_dayoung\anaconda3\lib\site-packages\lightgbm\sklearn.py:736: UserWarning: 'verbose' argument is deprecated and will be removed in a future release of LightGBM. Pass 'log_evaluation()' callback via 'callbacks' argument instead.
  _log_warning("'verbose' argument is deprecated and will be removed in a future release of LightGBM. "

[100]	valid_0's l2: 2.56673e+12
[200]	valid_0's l2: 2.45583e+12
[300]	valid_0's l2: 2.42575e+12
[400]	valid_0's l2: 2.43392e+12
[100]	valid_0's l2: 4.89194e+12
[200]	valid_0's l2: 4.40922e+12
[300]	valid_0's l2: 4.19146e+12
[400]	valid_0's l2: 4.05951e+12
[500]	valid_0's l2: 3.96931e+12
[600]	valid_0's l2: 3.91727e+12
[700]	valid_0's l2: 3.88462e+12
[800]	valid_0's l2: 3.87695e+12
[900]	valid_0's l2: 3.87088e+12
[100]	valid_0's l2: 3.14361e+12
[200]	valid_0's l2: 2.79286e+12
[300]	valid_0's l2: 2.59302e+12
[400]	valid_0's l2: 2.47608e+12
[500]	valid_0's l2: 2.40386e+12
[600]	valid_0's l2: 2.36407e+12
[700]	valid_0's l2: 2.38505e+12
[100]	valid_0's l2: 1.60592e+12
[200]	valid_0's l2: 1.40227e+12
[300]	valid_0's l2: 1.30053e+12
[400]	valid_0's l2: 1.25184e+12
[500]	valid_0's l2: 1.23543e+12
[600]	valid_0's l2: 1.23595e+12
[100]	valid_0's l2: 1.96107e+12
[200]	valid_0's l2: 1.75478e+12
[300]	valid_0's l2: 1.64513e+12
[400]	valid_0's l2: 1.58132e+12
[500]	valid_0's l2: 1.54801e+12
[600]	valid_0's l2: 1.52159e+12
[700]	valid_0's l2: 1.50655e+12
[800]	valid_0's l2: 1.49834e+12
[900]	valid_0's l2: 1.50018e+12

In [109]:

# 예측 결과값 정리

preds = []

for model in models:
    preds.append(model.predict(X_test))

In [110]:

# 최고 점수 출력
gs.best_score_

Out[110]:

-2334525343085.6494

In [112]:

# 예측 결과값 파일에 저장 하는 과정
gslgbm = submission.copy()
gslgbm['box_off_num'] = np.mean(preds, axis = 0)
gslgbm.to_csv('data/gslgbm.csv', index = False)

6. 결과 및 결언¶

Conclusion & Discussion¶

a. lightGBM (base model)¶

In [113]:

singleLGBM

Out[113]:

	title	box_off_num
0	용서는 없다	2817995.2
1	아빠가 여자를 좋아해	375377.2
2	하모니	-569324.3
3	의형제	1581189.0
4	평행 이론	-527780.6
...	...	...
238	해에게서 소년에게	500784.4
239	울보 권투부	1013858.4
240	어떤살인	1682067.7
241	말하지 못한 비밀	300216.3
242	조선안방 스캔들-칠거지악 2	11390.0

243 rows × 2 columns

b. k-fold lightGBM (k-fold model)¶

In [114]:

kfoldLightGBM

Out[114]:

	title	box_off_num
0	용서는 없다	3367422.1
1	아빠가 여자를 좋아해	961138.9
2	하모니	1097929.7
3	의형제	2097270.8
4	평행 이론	781476.3
...	...	...
238	해에게서 소년에게	177095.8
239	울보 권투부	210255.5
240	어떤살인	360547.6
241	말하지 못한 비밀	181508.2
242	조선안방 스캔들-칠거지악 2	55171.2

243 rows × 2 columns

c. feature engineering (fe)¶

In [115]:

feLightGBM

Out[115]:

	title	box_off_num
0	용서는 없다	3395492.7
1	아빠가 여자를 좋아해	823543.9
2	하모니	1162055.4
3	의형제	2184689.1
4	평행 이론	809328.8
...	...	...
238	해에게서 소년에게	81854.0
239	울보 권투부	54816.4
240	어떤살인	410490.0
241	말하지 못한 비밀	139172.4
242	조선안방 스캔들-칠거지악 2	28897.8

243 rows × 2 columns

d. grid search (hyperparameter tuning)¶

In [116]:

gslgbm

Out[116]:

	title	box_off_num
0	용서는 없다	2974959.7
1	아빠가 여자를 좋아해	982313.1
2	하모니	1283210.4
3	의형제	1681758.5
4	평행 이론	909584.5
...	...	...
238	해에게서 소년에게	78861.3
239	울보 권투부	127602.2
240	어떤살인	447047.3
241	말하지 못한 비밀	276243.3
242	조선안방 스캔들-칠거지악 2	58072.2

243 rows × 2 columns

블로그 업로드 용 코드¶

In [117]:

from IPython.core.display import display, HTML

display(HTML("<style>.container { width:90% !important; }</style>"))

제출 결과

'AI > Machine Learning' 카테고리의 다른 글

[Practical Time Series Analysis (실전 시계열 분석)] Chapter 06 시계열의 통계 모델 (0)	2022.03.23
[Practical Time Series Analysis (실전 시계열 분석)] Chapter 05 시간 데이터 저장 (0)	2022.03.23
[Practical Time Series Analysis (실전 시계열 분석)] Chapter 04 시계열 데이터의 시뮬레이션 (0)	2022.03.22
[Practical Time Series Analysis (실전 시계열 분석)] Chapter 03 시계열의 탐색적 자료 분석 (0)	2022.03.20
[DACON] 영화 데이터를 활용한 데이터 분석1 (EDA) (0)	2022.01.30

All-young's Blog

[DACON] 문화 | 영화 관객수 예측 모델 개발

영화 관객 수 데이터를 활용한 데이터 분석¶

1. 라이브러리 및 데이터¶

Library & Data¶

2. 탐색적 자료분석¶

Exploratory Data Analysis (EDA)¶

3. 데이터 전처리¶

Data Cleansing & Pre-Processing¶

4. 변수 선택 및 모델 구축¶

Feature Engineering & Initial Modeling¶

5. 모델 학습 및 검증¶

Model Tuning & Evaluation¶

a. lightGBM (base model)¶

b. k-fold lightGBM (k-fold model)¶

c. feature engineering (fe)¶

d. grid search (hyperparameter tuning)¶

LightGBM¶

k-fold 교차 검증 (k-fold cross validation)¶

feature engineering¶

Grid Search¶

가장 우수한 성능을 보이는 모델의 하이퍼 파라미터를 찾기 위해서¶

6. 결과 및 결언¶

Conclusion & Discussion¶

a. lightGBM (base model)¶

b. k-fold lightGBM (k-fold model)¶

c. feature engineering (fe)¶

d. grid search (hyperparameter tuning)¶

블로그 업로드 용 코드¶

제출 결과

'AI > Machine Learning' 카테고리의 다른 글

티스토리툴바

[DACON] 문화 | 영화 관객수 예측 모델 개발

영화 관객 수 데이터를 활용한 데이터 분석¶

1. 라이브러리 및 데이터¶

Library & Data¶

2. 탐색적 자료분석¶

Exploratory Data Analysis (EDA)¶

3. 데이터 전처리¶

Data Cleansing & Pre-Processing¶

4. 변수 선택 및 모델 구축¶

Feature Engineering & Initial Modeling¶

5. 모델 학습 및 검증¶

Model Tuning & Evaluation¶

a. lightGBM (base model)¶

b. k-fold lightGBM (k-fold model)¶

c. feature engineering (fe)¶

d. grid search (hyperparameter tuning)¶

LightGBM¶

k-fold 교차 검증 (k-fold cross validation)¶

feature engineering¶

Grid Search¶

가장 우수한 성능을 보이는 모델의 하이퍼 파라미터를 찾기 위해서¶

6. 결과 및 결언¶

Conclusion & Discussion¶

a. lightGBM (base model)¶

b. k-fold lightGBM (k-fold model)¶

c. feature engineering (fe)¶

d. grid search (hyperparameter tuning)¶

블로그 업로드 용 코드¶

제출 결과

'AI > Machine Learning' 카테고리의 다른 글

'AI/Machine Learning' Related Articles

티스토리툴바