Kaggle Spaceship Titanic
EDA & Preprocessing
데이터 셋
Kaggle Spaceship Titanic
사용 라이브러리
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
| import numpy as np
import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt
import koreanize_matplotlib
import random
import glob
from sklearn.metrics import accuracy_score
from sklearn.model_selection import train_test_split
from sklearn.ensemble import RandomForestClassifier
import xgboost as xgb
from xgboost import XGBClassifier
import lightgbm as lgbm
from lightgbm import LGBMClassifier
import warnings
warnings.filterwarnings("ignore")
|
Data Load
1
2
| path = glob.glob("data/*")
path
|
1
| ['data\\sample_submission.csv', 'data\\test.csv', 'data\\train.csv']
|
1
2
3
4
5
| train = pd.read_csv(path[2])
test = pd.read_csv(path[1])
sub = pd.read_csv(path[0])
train.shape, test.shape
|
1
| ((8693, 14), (4277, 13))
|
EDA
기본 정보
1
2
| display(train.head(3))
display(test.head(3))
|
| PassengerId | HomePlanet | CryoSleep | Cabin | Destination | Age | VIP | RoomService | FoodCourt | ShoppingMall | Spa | VRDeck | Name | Transported |
|---|
| 0 | 0001_01 | Europa | False | B/0/P | TRAPPIST-1e | 39.0 | False | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | Maham Ofracculy | False |
|---|
| 1 | 0002_01 | Earth | False | F/0/S | TRAPPIST-1e | 24.0 | False | 109.0 | 9.0 | 25.0 | 549.0 | 44.0 | Juanna Vines | True |
|---|
| 2 | 0003_01 | Europa | False | A/0/S | TRAPPIST-1e | 58.0 | True | 43.0 | 3576.0 | 0.0 | 6715.0 | 49.0 | Altark Susent | False |
|---|
| PassengerId | HomePlanet | CryoSleep | Cabin | Destination | Age | VIP | RoomService | FoodCourt | ShoppingMall | Spa | VRDeck | Name |
|---|
| 0 | 0013_01 | Earth | True | G/3/S | TRAPPIST-1e | 27.0 | False | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | Nelly Carsoning |
|---|
| 1 | 0018_01 | Earth | False | F/4/S | TRAPPIST-1e | 19.0 | False | 0.0 | 9.0 | 0.0 | 2823.0 | 0.0 | Lerome Peckers |
|---|
| 2 | 0019_01 | Europa | True | C/0/S | 55 Cancri e | 31.0 | False | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | Sabih Unhearfus |
|---|
PassengerID: ggg-pp 포멧, ggg는 함께 여행하는 그룹, pp는 그룹내의 번호HomePlanet: 출발 행성CryoSleep: 극저온 수면 여부, 극저온 수면 중인 경우 객실에 있음Cabin: deck/num/side 포멧, 객실 번호side: P=Port, S=Starboard
Destination: 도착지AgeVIPRoomService, FoodCourt, ShoppingMall, Spa, VRDeck: 편의시설 이용 여부 (요금 청구 내역)NameTransported: 다른 차원으로 이송되었는지 여부 (목표 값)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
| <class 'pandas.core.frame.DataFrame'>
RangeIndex: 8693 entries, 0 to 8692
Data columns (total 14 columns):
# Column Non-Null Count Dtype
--- ------ -------------- -----
0 PassengerId 8693 non-null object
1 HomePlanet 8492 non-null object
2 CryoSleep 8476 non-null object
3 Cabin 8494 non-null object
4 Destination 8511 non-null object
5 Age 8514 non-null float64
6 VIP 8490 non-null object
7 RoomService 8512 non-null float64
8 FoodCourt 8510 non-null float64
9 ShoppingMall 8485 non-null float64
10 Spa 8510 non-null float64
11 VRDeck 8505 non-null float64
12 Name 8493 non-null object
13 Transported 8693 non-null bool
dtypes: bool(1), float64(6), object(7)
memory usage: 891.5+ KB
|
| Age | RoomService | FoodCourt | ShoppingMall | Spa | VRDeck |
|---|
| count | 8514.000000 | 8512.000000 | 8510.000000 | 8485.000000 | 8510.000000 | 8505.000000 |
|---|
| mean | 28.827930 | 224.687617 | 458.077203 | 173.729169 | 311.138778 | 304.854791 |
|---|
| std | 14.489021 | 666.717663 | 1611.489240 | 604.696458 | 1136.705535 | 1145.717189 |
|---|
| min | 0.000000 | 0.000000 | 0.000000 | 0.000000 | 0.000000 | 0.000000 |
|---|
| 25% | 19.000000 | 0.000000 | 0.000000 | 0.000000 | 0.000000 | 0.000000 |
|---|
| 50% | 27.000000 | 0.000000 | 0.000000 | 0.000000 | 0.000000 | 0.000000 |
|---|
| 75% | 38.000000 | 47.000000 | 76.000000 | 27.000000 | 59.000000 | 46.000000 |
|---|
| max | 79.000000 | 14327.000000 | 29813.000000 | 23492.000000 | 22408.000000 | 24133.000000 |
|---|
1
| _ = train.hist(bins=50, figsize=(12, 8))
|
결측치 확인
1
2
3
4
| fig, ax = plt.subplots(1, 2, figsize=(12, 7))
sns.heatmap(train.isnull(), ax=ax[0]).set_title("Train-Missing")
sns.heatmap(test.isnull(), ax=ax[1]).set_title("Test-Missing")
plt.show()
|
1
2
3
4
5
| temp_train = train.isnull().mean()*100
temp_test = test.isnull().mean()*100
display(temp_train)
display(temp_test)
|
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
| PassengerId 0.000000
HomePlanet 2.312205
CryoSleep 2.496261
Cabin 2.289198
Destination 2.093639
Age 2.059128
VIP 2.335212
RoomService 2.082135
FoodCourt 2.105142
ShoppingMall 2.392730
Spa 2.105142
VRDeck 2.162660
Name 2.300702
Transported 0.000000
dtype: float64
PassengerId 0.000000
HomePlanet 2.034136
CryoSleep 2.174421
Cabin 2.338087
Destination 2.151040
Age 2.127660
VIP 2.174421
RoomService 1.917232
FoodCourt 2.478373
ShoppingMall 2.291326
Spa 2.361468
VRDeck 1.870470
Name 2.197802
dtype: float64
|
대략 2% 정도의 결측치가 존대함
상관 관계
결측치 대체를 상관 관계를 이용해 처리할 예정
1
2
| plt.figure(figsize=(12, 8))
_ = sns.heatmap(train.corr(), cmap="coolwarm", annot=True, mask=np.triu(np.ones_like(train.corr()))).set_title("Corr")
|
PassengerID
ggg의 경우 함께 여행하는 그룹을 의미하므로, 동반자 여부를 확인할 수 있을거라 생각됨
pp는 개인 번호이므로 크게 의미가 없을것이라 생각
1
2
| train["ggg"] = train["PassengerId"].str.split("_", expand=True)[0]
test["ggg"] = test["PassengerId"].str.split("_", expand=True)[0]
|
1
| (train["ggg"].value_counts()>1).sum()
|
1
| (train["ggg"].value_counts()>1).sum() / train["ggg"].count() * 100
|
1
| (test["ggg"].value_counts()>1).sum() / test["ggg"].count() * 100
|
1412명, 약 16% 정도는 일행이 있음
이를 바탕으로, 일행 수를 나타내는 파생 변수를 생성
1
2
3
4
| ggg_num_train, ggg_num_test = train["ggg"].value_counts().sort_index(), test["ggg"].value_counts().sort_index()
train["ggg"] = train["ggg"].apply(lambda x: ggg_num_train[x])
test["ggg"] = train["ggg"].apply(lambda x: ggg_num_test[x])
|
1
2
| display(train["ggg"].sample(3))
display(test["ggg"].sample(3))
|
1
2
3
4
5
6
7
8
9
10
11
| 7004 1
207 7
2897 3
Name: ggg, dtype: int64
2134 1
678 1
2535 1
Name: ggg, dtype: int64
|
HomePlanet
1
| train["HomePlanet"].isnull().sum(), test["HomePlanet"].isnull().sum()
|
1
2
3
4
| fig, ax =plt.subplots(1, 2, figsize=(12, 5))
sns.countplot(x=train["HomePlanet"].sort_values(), ax=ax[0]).set_title("Train")
sns.countplot(x=test["HomePlanet"].sort_values(), ax=ax[1]).set_title("Test")
plt.show()
|
1
| _ = sns.countplot(data=train, x="HomePlanet", hue="Transported")
|
결측치를 랜덤하게 설정
1
2
| train["HomePlanet"].fillna(random.choice(train["HomePlanet"].unique()), inplace=True)
test["HomePlanet"].fillna(random.choice(test["HomePlanet"].unique()), inplace=True)
|
1
| train["HomePlanet"].isnull().sum(), test["HomePlanet"].isnull().sum()
|
Destination
1
| train["Destination"].isnull().sum(), test["Destination"].isnull().sum()
|
1
2
3
4
| fig, ax =plt.subplots(1, 2, figsize=(12, 5))
sns.countplot(x=train["Destination"].sort_values(), ax=ax[0]).set_title("Train")
sns.countplot(x=test["Destination"].sort_values(), ax=ax[1]).set_title("Test")
plt.show()
|
1
| _ = sns.countplot(data=train, x="Destination", hue="Transported")
|
결측치를 랜덤하게 설정
1
2
| train["Destination"].fillna(random.choice(train["Destination"].unique()), inplace=True)
test["Destination"].fillna(random.choice(test["Destination"].unique()), inplace=True)
|
1
| train["Destination"].isnull().sum(), test["Destination"].isnull().sum()
|
CryoSleep
1
| train["CryoSleep"].isnull().sum(), test["CryoSleep"].isnull().sum()
|
1
2
3
4
| fig, ax =plt.subplots(1, 2, figsize=(12, 5))
sns.countplot(x=train["CryoSleep"].sort_values(), ax=ax[0]).set_title("Train")
sns.countplot(x=test["CryoSleep"].sort_values(), ax=ax[1]).set_title("Test")
plt.show()
|
극저온 수면을 취하는 사람들은 전부 사망하지 않았을까?
1
| _ = sns.countplot(data=train, x="CryoSleep", hue="Transported")
|
1
2
3
4
5
| data = train.groupby("CryoSleep")["Transported"].count().values.tolist()
labels = train.groupby("CryoSleep")["Transported"].count().index.tolist()
colors = sns.color_palette('pastel')[0:2]
_ = plt.pie(data, labels=labels, colors=colors, autopct="%.2f%%")
|
8500여 명 중, 3000(36%)명 정도가 극저온 수면을 선택함
1
2
3
4
5
6
| # 극저온 수면을하지 않은 그룹
data = train.groupby("CryoSleep")["Transported"].value_counts()[0].values.tolist()
labels = train.groupby("CryoSleep")["Transported"].value_counts()[0].index.tolist()
colors = sns.color_palette('pastel')[2:4]
_ = plt.pie(data, labels=labels, colors=colors, autopct="%.2f%%")
|
극저온 수면을하지 않은 사람들의 약 67%는 생존했음
1
2
3
4
5
6
| # 극저운 수면을한 그룹
data = train.groupby("CryoSleep")["Transported"].value_counts()[1].values.tolist()
labels = train.groupby("CryoSleep")["Transported"].value_counts()[1].index.tolist()
colors = sns.color_palette('pastel')[4:6]
_ = plt.pie(data, labels=labels, colors=colors, autopct="%.2f%%")
|
극저온 수면을 한 사람들의 약 82%는 사망했음
1
| _ = train.groupby("CryoSleep")["Transported"].value_counts().plot.bar(rot=0)
|
극저온 수면을 선택한 사람들의 사망 비율이 높다는 것을 확인 할 수 있음
train의 CryoSleep의 결측치는 해당 정보를 이용해 보완해 줄 수 있을 것 같음
극저온 수면을 선택한 사람들의 출발/도착 행성은?
1
| _ = sns.countplot(data=train, x="CryoSleep", hue="HomePlanet").set_title("CryoSleep - HomePlanet")
|
1
| _ = sns.countplot(data=train, x="CryoSleep", hue="Destination").set_title("CryoSleep - Destination")
|
출발지/도착지와 극저온 수면 선택 여부는 크게 상관이 없어보임
단순히, 지구 출발자와 TRAPPIST-1e가 도착지인 사람들이 많이 선택했음
1
| train.groupby("Destination")["CryoSleep"].value_counts(normalize=True)*100
|
1
2
3
4
5
6
7
8
| Destination CryoSleep
55 Cancri e False 57.915718
True 42.084282
PSO J318.5-22 False 50.707851
True 49.292149
TRAPPIST-1e False 67.777217
True 32.222783
Name: CryoSleep, dtype: float64
|
위 확률을 이용해 랜덤하게 부여할 예정
거리와 상관있지 않을까?
출발지는 Earth, Mars, Europa,
도착지는 TRAPPIST-1e, PSO J318.5-22, 55 Cancri e
Mars는 지구로부터 약 0.000042 광년 떨어져 있음Europa는 지구로부터 약 0.000066 광년 떨어져 있음TRAPPIST-1e는 지구로부터 약 39.6 광년 떨어져 있음PSO J318.5-22는 지구로부터 약 80광년 떨어져 있음55 Cancri e는 지구로부터 약 40.9광년 떨어져 있음
사실상 목저지만 이용해도 될 것 같음
위, Destination 파트에서 봤듯이 Transported와 크게 상관이 없는 것으로 보임
train의 CryoSleep은 Transported를 바탕으로 결측치를 채우고,
test의 CryoSleep은 Destination을 바탕으로 채워줌
1
2
3
4
5
6
7
8
| # train
def base_Transported(cols):
Cryosleep, Transported = cols[0], cols[1]
# Cryosleep이 결측치면,
if pd.isnull(Cryosleep):
if Transported: return True # 사망했다면 True
else: return False # 사망하지 않았다면 False
else: return Cryosleep # 결측치가 아닌 경우 그대로 반환
|
1
| train["CryoSleep"] = train[["CryoSleep", "Transported"]].apply(base_Transported, axis=1)
|
1
2
3
4
5
6
7
8
9
10
11
| # test
def base_Destination_proba(cols):
Cryosleep, Destination = cols[0], cols[1]
if pd.isnull(Cryosleep):
if Destination=="TRAPPIST-1e":
return random.choices([True, False], weights=[0.33, 0.67])[-1]
elif Destination=="55 Cancri e":
return random.choices([True, False], weights=[0.41, 0.59])[-1]
else:
return random.choices([True, False], weights=[0.49, 0.51])[-1]
else: return Cryosleep
|
1
| test["CryoSleep"] = test[["CryoSleep", "Destination"]].apply(base_Destination_proba, axis=1)
|
1
| train["CryoSleep"].isnull().sum(), test["CryoSleep"].isnull().sum()
|
Cabin
deck/num/side 형식을 가지고 있음
1
| train["Cabin"].isnull().sum(), test["Cabin"].isnull().sum()
|
1
| train[train["Cabin"].duplicated()].iloc[:5]
|
| PassengerId | HomePlanet | CryoSleep | Cabin | Destination | Age | VIP | RoomService | FoodCourt | ShoppingMall | Spa | VRDeck | Name | Transported | ggg |
|---|
| 3 | 0003_02 | Europa | False | A/0/S | TRAPPIST-1e | 33.0 | False | 0.0 | 1283.0 | 371.0 | 3329.0 | 193.0 | Solam Susent | False | 2 |
|---|
| 10 | 0008_02 | Europa | True | B/1/P | TRAPPIST-1e | 34.0 | False | 0.0 | 0.0 | NaN | 0.0 | 0.0 | Altardr Flatic | True | 3 |
|---|
| 11 | 0008_03 | Europa | False | B/1/P | 55 Cancri e | 45.0 | False | 39.0 | 7295.0 | 589.0 | 110.0 | 124.0 | Wezena Flatic | True | 3 |
|---|
| 22 | 0020_02 | Earth | True | E/0/S | 55 Cancri e | 49.0 | False | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | Glendy Brantuarez | False | 6 |
|---|
| 23 | 0020_03 | Earth | True | E/0/S | 55 Cancri e | 29.0 | False | 0.0 | 0.0 | NaN | 0.0 | 0.0 | Mollen Mcfaddennon | False | 6 |
|---|
같은 객실을 사용한 사람들은 같은 그룹에 속해있는 사람임
deck과 side 정보만 파생변수로 만들어 사용함
결측치 처리를 위해 결측치가 없는 데이터만 일부 먼저 사용
1
2
| temp = train[~train["Cabin"].isnull()]
temp.shape
|
1
2
| temp["Deck"] = temp["Cabin"].apply(lambda x: x.split("/")[0])
temp["Side"] = temp["Cabin"].apply(lambda x: x.split("/")[-1])
|
1
2
3
4
| fig, ax = plt.subplots(1, 2, figsize=(12, 6))
sns.countplot(x=temp["Deck"], ax=ax[0]).set_title("Deck")
sns.countplot(x=temp["Side"], ax=ax[1]).set_title("Side")
plt.show()
|
Deck과 달리, Side는 거의 균일하게 분포되어 있음
규모가 다를까라는 의문이 생김
1
| _ = sns.countplot(data=temp, x="Deck", hue="Side")
|
Side에 따라 Deck도 비교적 균일하게 분포되어있음 -> 규모도 거의 동일
1
| temp.groupby("Transported")["Deck"].value_counts(normalize=True)*100
|
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
| Transported Deck
False F 37.120493
G 29.364326
E 13.353890
D 6.427894
C 5.668880
B 4.909867
A 3.059772
T 0.094877
True G 30.878915
F 28.728378
B 13.370734
C 11.874708
E 7.316503
D 4.838710
A 2.968677
T 0.023375
Name: Deck, dtype: float64
|
1
| temp.groupby("Transported")["Side"].value_counts(normalize=True)*100
|
1
2
3
4
5
6
| Transported Side
False P 54.743833
S 45.256167
True S 55.633474
P 44.366526
Name: Side, dtype: float64
|
1
2
3
4
5
6
7
8
9
10
11
12
13
| # train
def base_Transported_proba(cols):
Cabin, Transported = cols[0], cols[1]
if pd.isnull(Cabin):
if Transported:
Deck = random.choices(["G", "F", "B", "C", "E", "D", "A", "T"], weights=[0.31, 0.29, 0.13, 0.12, 0.07, 0.05, 0.03, 0.002])[0]
Side = random.choices(["P", "S"], weights=[0.56, 0.44])[0]
return f"{Deck}/{Side}"
else:
Deck = random.choices(["F", "G", "E", "D", "C", "B", "A", "T"], weights=[0.37, 0.29, 0.13, 0.06, 0.05, 0.04, 0.03, 0.001])[0]
Side = random.choices(["P", "S"], weights=[0.55, 0.45])[0]
return f"{Deck}/{Side}"
else: return Cabin
|
1
| train["Cabin"] = train[["Cabin", "Transported"]].apply(base_Transported_proba, axis=1)
|
train의 Cabin은 위 2개의 표를 이용해 랜덤하게 채워줌
1
| temp["Deck"].value_counts(normalize=True)*100
|
1
2
3
4
5
6
7
8
9
| F 32.893807
G 30.127149
E 10.313162
B 9.171180
C 8.794443
D 5.627502
A 3.013892
T 0.058865
Name: Deck, dtype: float64
|
1
| temp["Side"].value_counts(normalize=True)*100
|
1
2
3
| S 50.482694
P 49.517306
Name: Side, dtype: float64
|
1
2
3
| # test
test["Cabin"].fillna(str(random.choices(["F", "G", "E", "B", "C", "D", "A", "T"], weights=[0.33, 0.30, 0.1, 0.09, 0.09, 0.06, 0.03, 0.0006])[0]) \
+ "/" + str(random.choices(["S", "P"], weights=[0.5, 0.5])[0]), inplace=True)
|
1
| train["Cabin"].isnull().sum(), test["Cabin"].isnull().sum()
|
test의 Cabin은 위 2개의 표를 이용해 랜덤하게 채워줌
1
2
3
4
5
| train["Deck"] = train["Cabin"].apply(lambda x: x.split("/")[0])
train["Side"] = train["Cabin"].apply(lambda x: x.split("/")[-1])
test["Deck"] = test["Cabin"].apply(lambda x: x.split("/")[0])
test["Side"] = test["Cabin"].apply(lambda x: x.split("/")[-1])
|
Age
1
| train["Age"].isnull().sum(), test["Age"].isnull().sum()
|
1
2
3
4
| fig, ax = plt.subplots(1, 2, figsize=(12, 6))
sns.kdeplot(data=train, x="Age", hue="VIP", shade=True, ax=ax[0]).set_title("Age - Train")
sns.kdeplot(data=test, x="Age", hue="VIP", shade=True, ax=ax[1]).set_title("Age - Test")
plt.show()
|
1
| train.groupby("VIP")["Age"].describe()
|
| count | mean | std | min | 25% | 50% | 75% | max |
|---|
| VIP | | | | | | | | |
|---|
| False | 8119.0 | 28.639611 | 14.469895 | 0.0 | 19.0 | 27.0 | 38.0 | 79.0 |
|---|
| True | 198.0 | 37.449495 | 11.611957 | 18.0 | 29.0 | 34.0 | 44.0 | 73.0 |
|---|
1
| test.groupby("VIP")["Age"].describe()
|
| count | mean | std | min | 25% | 50% | 75% | max |
|---|
| VIP | | | | | | | | |
|---|
| False | 4023.0 | 28.487447 | 14.238827 | 0.0 | 19.0 | 26.0 | 37.0 | 79.0 |
|---|
| True | 73.0 | 34.534247 | 9.771106 | 18.0 | 27.0 | 32.0 | 39.0 | 67.0 |
|---|
VIP 여부를 이용해 결측치를 채워줌
1
2
3
4
5
6
| def age_by_VIP(cols):
age, vip = cols[0], cols[1]
if pd.isna(age):
if vip: return 35
else: return 28
else: return age
|
1
2
| train["Age"] = train[["Age", "VIP"]].apply(age_by_VIP, axis=1)
test["Age"] = test[["Age", "VIP"]].apply(age_by_VIP, axis=1)
|
1
| train["Age"].isnull().sum(), test["Age"].isnull().sum()
|
band_age 컬럼을 생성
1
2
| train["Band_Age"] = train["Age"].apply(lambda x: int(str(x)[0]+"0"))
test["Band_Age"] = test["Age"].apply(lambda x: int(str(x)[0]+"0"))
|
1
| _ = sns.countplot(data=train, x="Band_Age", hue="Transported").set_title("Transported - Age Band")
|
20대 미만 -> child
20~40대 -> adult
40~60대 -> middle
60대~ -> old
1
2
3
4
5
| def age_str(col):
if col in [0, 10]: return "child"
elif col in [20, 40]: return "adult"
elif col in [50, 60]: return "middle"
else: return "old"
|
1
2
| train["Band_Age"] = train["Band_Age"].apply(age_str)
test["Band_Age"] = test["Band_Age"].apply(age_str)
|
VIP
1
| train["VIP"].isnull().sum(), test["VIP"].isnull().sum()
|
adult, middle이면 True, 아니면 False로 결측치를 대체함
1
2
3
4
5
6
7
| def base_band_VIP(cols):
band, vip = cols[0], cols[1]
if pd.isna(vip):
if band in ["adult", "middle"]: return True
else: return False
else: return vip
|
1
2
| train["VIP"] = train[["Band_Age", "VIP"]].apply(base_band_VIP, axis=1)
test["VIP"] = test[["Band_Age", "VIP"]].apply(base_band_VIP, axis=1)
|
1
| train["VIP"].isnull().sum(), test["VIP"].isnull().sum()
|
실수형 변수
1
| nums = ["RoomService", "FoodCourt", "ShoppingMall", "Spa", "VRDeck"]
|
1
| train[nums].isnull().sum()
|
1
2
3
4
5
6
| RoomService 181
FoodCourt 183
ShoppingMall 208
Spa 183
VRDeck 188
dtype: int64
|
1
| test[nums].isnull().sum()
|
1
2
3
4
5
6
| RoomService 82
FoodCourt 106
ShoppingMall 98
Spa 101
VRDeck 80
dtype: int64
|
1
| _ = train[nums].hist(bins=30, figsize=(14, 7))
|
이용 여부로 변경해도 좋을꺼 같다는 생각을 함
1
2
| train[nums] = train[nums].fillna(0)
test[nums] = test[nums].fillna(0)
|
1
2
| train["Service"] = train[nums].sum(axis=1)
test["Service"] = test[nums].sum(axis=1)
|
1
2
| train["Service"] = train["Service"].apply(lambda x: False if x==0 else True)
test["Service"] = test["Service"].apply(lambda x: False if x==0 else True)
|
| PassengerId | HomePlanet | CryoSleep | Cabin | Destination | Age | VIP | RoomService | FoodCourt | ShoppingMall | Spa | VRDeck | Name | Transported | ggg | Deck | Side | Band_Age | Service |
|---|
| 0 | 0001_01 | Europa | False | B/0/P | TRAPPIST-1e | 39.0 | False | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | Maham Ofracculy | False | 1 | B | P | old | False |
|---|
| 1 | 0002_01 | Earth | False | F/0/S | TRAPPIST-1e | 24.0 | False | 109.0 | 9.0 | 25.0 | 549.0 | 44.0 | Juanna Vines | True | 1 | F | S | adult | True |
|---|
| 2 | 0003_01 | Europa | False | A/0/S | TRAPPIST-1e | 58.0 | True | 43.0 | 3576.0 | 0.0 | 6715.0 | 49.0 | Altark Susent | False | 2 | A | S | middle | True |
|---|
| 3 | 0003_02 | Europa | False | A/0/S | TRAPPIST-1e | 33.0 | False | 0.0 | 1283.0 | 371.0 | 3329.0 | 193.0 | Solam Susent | False | 2 | A | S | old | True |
|---|
| 4 | 0004_01 | Earth | False | F/1/S | TRAPPIST-1e | 16.0 | False | 303.0 | 70.0 | 151.0 | 565.0 | 2.0 | Willy Santantines | True | 1 | F | S | child | True |
|---|
불필요한 열 삭제
1
2
3
| drop_list = ["PassengerId", "Cabin", "Age", "Name"]
drop_list.extend(nums)
drop_list
|
1
2
3
4
5
6
7
8
9
| ['PassengerId',
'Cabin',
'Age',
'Name',
'RoomService',
'FoodCourt',
'ShoppingMall',
'Spa',
'VRDeck']
|
1
2
| train.drop(columns=drop_list, axis=1, inplace=True)
test.drop(columns=drop_list, axis=1, inplace=True)
|
1
2
| train.to_csv("data/pre_train.csv", index=False)
test.to_csv("data/pre_test.csv", index=False)
|
Model - Classifier
Data Load
1
2
| path = glob.glob("data/*")
path
|
1
2
3
4
5
| ['data\\pre_test.csv',
'data\\pre_train.csv',
'data\\sample_submission.csv',
'data\\test.csv',
'data\\train.csv']
|
1
2
3
4
5
| train = pd.read_csv(path[1])
test = pd.read_csv(path[0])
sub = pd.read_csv(path[2])
train.shape, test.shape, sub.shape
|
1
| ((8693, 10), (4277, 9), (4277, 2))
|
인코딩
1
2
| display(train.dtypes)
display(test.dtypes)
|
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
| HomePlanet object
CryoSleep bool
Destination object
VIP bool
Transported bool
ggg int64
Deck object
Side object
Band_Age object
Service bool
dtype: object
HomePlanet object
CryoSleep bool
Destination object
VIP bool
ggg int64
Deck object
Side object
Band_Age object
Service bool
dtype: object
|
1
2
3
4
5
6
7
| encode_list = train.columns.tolist()
encode_list.remove("Transported")
encode_list.remove("ggg")
encode_list.remove("CryoSleep")
encode_list.remove("VIP")
encode_list.remove("Service")
encode_list
|
1
| ['HomePlanet', 'Destination', 'Deck', 'Side', 'Band_Age']
|
1
2
3
4
5
6
7
8
| temp_train, temp_test = [], []
for col in encode_list:
temp_train.append(pd.get_dummies(train[col], drop_first=True))
temp_test.append(pd.get_dummies(test[col], drop_first=True))
temp_train = pd.concat(temp_train, axis=1)
temp_test = pd.concat(temp_test, axis=1)
|
1
2
| train = pd.concat([train.drop(columns=encode_list, axis=1), temp_train], axis=1)
test = pd.concat([test.drop(columns=encode_list, axis=1), temp_test], axis=1)
|
1
| train.shape, test.shape
|
1
| ((8693, 20), (4277, 19))
|
Train
1
2
| label = "Transported"
features = test.columns.tolist()
|
1
2
3
| X_train, X_test, y_train, y_test = train_test_split(train[features], train[label], test_size=0.2, stratify=train[label])
print(f"X_train: {X_train.shape}\ny_train: {y_train.shape}\nX_test: {X_test.shape}\ny_test: {y_test.shape}")
|
1
2
3
4
| X_train: (6954, 19)
y_train: (6954,)
X_test: (1739, 19)
y_test: (1739,)
|
Random Forest
1
2
3
4
5
| clf_rf = RandomForestClassifier()
pred_rf = clf_rf.fit(X_train, y_train).predict(X_test)
accuracy_score(y_test, pred_rf)
|
1
| _ = sns.barplot(x=clf_rf.feature_importances_, y=features).set_title("RF")
|
XGBoost
1
2
3
4
5
| clf_xgb = XGBClassifier()
pred_xgb = clf_xgb.fit(X_train, y_train).predict(X_test)
accuracy_score(y_test, pred_rf)
|
1
| _ = xgb.plot_importance(clf_xgb)
|
LGBM
1
2
3
4
5
| clf_lgbm = LGBMClassifier()
pred_lgbm = clf_lgbm.fit(X_train, y_train).predict(X_test)
accuracy_score(y_test, pred_lgbm)
|
1
| _ = lgbm.plot_importance(clf_lgbm)
|
마무리
EDA와 전처리를 굉장히 만족스럽게했지만, 기대했던 성능에 미치지는 못했다..
베스트 스코어는 역시 lgbm을 사용한 경우로, 74점정도임
Comments powered by Disqus.