数据链接：https://pan.baidu.com/s/1sZ4tC8dk4jMTBgheXzbhcg
提取码：3qev
复制这段内容后打开百度网盘手机App，操作更方便哦

import numpy as np
import pandas as pd

# 展示
import matplotlib.pyplot as plt
import seaborn as sns
from pdpbox import pdp
# Sklearn
from sklearn.ensemble import RandomForestRegressor
from sklearn.model_selection import train_test_split
from IPython.display import display
from sklearn import metrics

• dbnos-被击倒的敌方玩家数量。
• 助攻-被队友击伤的敌方球员数量。
• 助推器-使用的助推器项目数。
• 损坏全部-造成的全部损坏。注：减去自身造成的伤害。
• headshotkills-用headshots杀死的敌方玩家数量。
• 治疗-使用的治疗项目数。
• ID-玩家ID
• 杀戮地点-在杀死的敌方玩家数量的比赛中排名。
• 杀戮点-基于玩家外部排名的杀戮。（把这看作是一个只有杀戮才重要的ELO排名。）如果rankpoints中有-1以外的值，那么任何0中的杀戮都应被视为“无”。
• 杀戮条纹-在短时间内杀死的敌人玩家的最大数量。
• 杀戮-被杀的敌方玩家数量。
• longestkill-死亡时玩家与被杀玩家之间的最长距离。这可能会误导玩家，因为击倒一名玩家并将其驱走可能会导致一个巨大的longestkill属性。
• matchDuration-匹配持续时间（秒）。
• matchID—标识匹配的ID。训练和测试集中都没有匹配项。
• matchType—标识数据来源的游戏模式的字符串。标准模式为“单人”、“双人”、“Squad”、“单人fpp”、“双人fpp”和“Squad fpp”；其他模式来自活动或自定义比赛。
• rankpoints-类似于ELO的玩家排名。这个排名是不一致的，在API的下一个版本中被弃用，因此请谨慎使用。值-1替换为“无”。
• 复活-这个球员复活队友的次数。
• 行驶距离-以米为单位的车辆总行驶距离。
• 道路撞车-车辆内的撞车次数。
• 游泳距离-以米为单位的游泳总距离。
• TeamKills-该玩家杀死队友的次数。
• 车辆类别-被摧毁的车辆数量。
• 步行距离-步行总距离，以米为单位。
• 所需武器-所携带武器的数量。
• Winpoints-基于Win的玩家外部排名。（把这看作是一个ELO排名，只有赢才重要。）如果rankpoints中有-1以外的值，那么winpoints中的任何0都应被视为“无”。
• group id-标识匹配中的组的ID。如果同一组玩家在不同的比赛中玩，他们每次都会有不同的groupid。
• numgroups-我们在匹配中有数据的组数。
• MaxPlace-我们在匹配中有数据的最差位置。这可能与numgroups不匹配，因为有时数据跳过放置。
• winplaceperc-预测的目标。这是一个百分位数的获胜位置，1代表第一名，0代表比赛的最后一名。它是根据maxplace而不是numgroups计算的，因此在匹配中可能有丢失的块。

train = pd.read_csv('train_V2.csv')
test = pd.read_csv('test_V2.csv')

train.head()

	Id	groupId	matchId	assists	boosts	damageDealt	DBNOs	headshotKills	heals	killPlace	...	revives	rideDistance	roadKills	swimDistance	teamKills	vehicleDestroys	walkDistance	weaponsAcquired	winPoints	winPlacePerc
0	7f96b2f878858a	4d4b580de459be	a10357fd1a4a91	0	0	0.00	0	0	0	60	...	0	0.0000	0	0.00	0	0	244.80	1	1466	0.4444
1	eef90569b9d03c	684d5656442f9e	aeb375fc57110c	0	0	91.47	0	0	0	57	...	0	0.0045	0	11.04	0	0	1434.00	5	0	0.6400
2	1eaf90ac73de72	6a4a42c3245a74	110163d8bb94ae	1	0	68.00	0	0	0	47	...	0	0.0000	0	0.00	0	0	161.80	2	0	0.7755
3	4616d365dd2853	a930a9c79cd721	f1f1f4ef412d7e	0	0	32.90	0	0	0	75	...	0	0.0000	0	0.00	0	0	202.70	3	0	0.1667
4	315c96c26c9aac	de04010b3458dd	6dc8ff871e21e6	0	0	100.00	0	0	0	45	...	0	0.0000	0	0.00	0	0	49.75	2	0	0.1875

5 rows × 29 columns

train.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 4446966 entries, 0 to 4446965
Data columns (total 29 columns):
Id                 object
groupId            object
matchId            object
assists            int64
boosts             int64
damageDealt        float64
DBNOs              int64
headshotKills      int64
heals              int64
killPlace          int64
killPoints         int64
kills              int64
killStreaks        int64
longestKill        float64
matchDuration      int64
matchType          object
maxPlace           int64
numGroups          int64
rankPoints         int64
revives            int64
rideDistance       float64
roadKills          int64
swimDistance       float64
teamKills          int64
vehicleDestroys    int64
walkDistance       float64
weaponsAcquired    int64
winPoints          int64
winPlacePerc       float64
dtypes: float64(6), int64(19), object(4)
memory usage: 983.9+ MB

train[train['winPlacePerc'].isnull()]

	Id	groupId	matchId	assists	boosts	damageDealt	DBNOs	headshotKills	heals	killPlace	...	revives	rideDistance	roadKills	swimDistance	teamKills	vehicleDestroys	walkDistance	weaponsAcquired	winPoints	winPlacePerc
2744604	f70c74418bb064	12dfbede33f92b	224a123c53e008	0	0	0.0	0	0	0	1	...	0	0.0	0	0.0	0	0	0.0	0	0	NaN

1 rows × 29 columns

train.drop(2744604,inplace=True)

train[train['winPlacePerc'].isnull()]

	Id	groupId	matchId	assists	boosts	damageDealt	DBNOs	headshotKills	heals	killPlace	...	revives	rideDistance	roadKills	swimDistance	teamKills	vehicleDestroys	walkDistance	weaponsAcquired	winPoints	winPlacePerc

0 rows × 29 columns

train['playersJoined'] = train.groupby('matchId')['matchId'].transform('count')
plt.figure(figsize=(15,10))
sns.countplot(train[train['playersJoined']>=75]['playersJoined'])
plt.show()

train.head()

	Id	groupId	matchId	assists	boosts	damageDealt	DBNOs	headshotKills	heals	killPlace	...	rideDistance	roadKills	swimDistance	teamKills	vehicleDestroys	walkDistance	weaponsAcquired	winPoints	winPlacePerc	playersJoined
0	7f96b2f878858a	4d4b580de459be	a10357fd1a4a91	0	0	0.00	0	0	0	60	...	0.0000	0	0.00	0	0	244.80	1	1466	0.4444	96
1	eef90569b9d03c	684d5656442f9e	aeb375fc57110c	0	0	91.47	0	0	0	57	...	0.0045	0	11.04	0	0	1434.00	5	0	0.6400	91
2	1eaf90ac73de72	6a4a42c3245a74	110163d8bb94ae	1	0	68.00	0	0	0	47	...	0.0000	0	0.00	0	0	161.80	2	0	0.7755	98
3	4616d365dd2853	a930a9c79cd721	f1f1f4ef412d7e	0	0	32.90	0	0	0	75	...	0.0000	0	0.00	0	0	202.70	3	0	0.1667	91
4	315c96c26c9aac	de04010b3458dd	6dc8ff871e21e6	0	0	100.00	0	0	0	45	...	0.0000	0	0.00	0	0	49.75	2	0	0.1875	97

5 rows × 30 columns

train['killsNorm'] = train['kills']*((100-train['playersJoined'])/100+1)
train['damageDealtNorm'] = train['damageDealt']*((100-train['playersJoined'])/100+1)
train['matchDurationNorm'] = train['matchDuration']*((100-train['playersJoined'])/100+1)
to_show = ['Id', 'kills','killsNorm','damageDealt', 'damageDealtNorm', 'matchDuration', 'matchDurationNorm']
train[to_show][:11]

	Id	kills	killsNorm	damageDealt	damageDealtNorm	matchDuration	matchDurationNorm
0	7f96b2f878858a	0	0.00	0.000	0.00000	1306	1358.24
1	eef90569b9d03c	0	0.00	91.470	99.70230	1777	1936.93
2	1eaf90ac73de72	0	0.00	68.000	69.36000	1318	1344.36
3	4616d365dd2853	0	0.00	32.900	35.86100	1436	1565.24
4	315c96c26c9aac	1	1.03	100.000	103.00000	1424	1466.72
5	ff79c12f326506	1	1.05	100.000	105.00000	1395	1464.75
6	95959be0e21ca3	0	0.00	0.000	0.00000	1316	1355.48
7	311b84c6ff4390	0	0.00	8.538	8.87952	1967	2045.68
8	1a68204ccf9891	0	0.00	51.600	53.14800	1375	1416.25
9	e5bb5a43587253	0	0.00	37.270	38.38810	1930	1987.90
10	2b574d43972813	0	0.00	28.380	28.66380	1811	1829.11

对于开挂的，应当给他们剔除掉

train['totalDistance'] = train['rideDistance'] + train['walkDistance'] + train['swimDistance']
train['killWithoutMoving'] = ((train['kills']>0)&(train['totalDistance']==0))

train['headshot_rate'] = train['headshotKills']/train['kills']
train['headshot_rate'] = train['headshot_rate'].fillna(0)

train[train['killWithoutMoving']==True].shape

(1535, 36)

train.drop(train[train['killWithoutMoving']==True].index,inplace=True)

train[train['roadKills']>1].shape

(1749, 36)

train.drop(train[train['roadKills']>10].index,inplace=True)

plt.figure(figsize=(15,10))
sns.countplot(data=train,x=train['kills']).set_title('Kills')
plt.show()

train[train['kills']>10].shape

(8120, 36)

train.drop(train[train['kills']>30].index,inplace=True)

plt.figure(figsize=(15,10))
sns.distplot(train['headshot_rate'],bins=10)
plt.show()

<Figure size 1080x720 with 0 Axes>

categorical变量

train['matchType'].nunique()

16

train = pd.get_dummies(train,columns=['matchType'])
train.head()

	Id	groupId	matchId	assists	boosts	damageDealt	DBNOs	headshotKills	heals	killPlace	...	matchType_normal-duo	matchType_normal-duo-fpp	matchType_normal-solo	matchType_normal-solo-fpp	matchType_normal-squad	matchType_normal-squad-fpp	matchType_solo	matchType_solo-fpp	matchType_squad	matchType_squad-fpp
0	7f96b2f878858a	4d4b580de459be	a10357fd1a4a91	0	0	0.00	0	0	0	60	...	0	0	0	0	0	0	0	0	0	1
1	eef90569b9d03c	684d5656442f9e	aeb375fc57110c	0	0	91.47	0	0	0	57	...	0	0	0	0	0	0	0	0	0	1
2	1eaf90ac73de72	6a4a42c3245a74	110163d8bb94ae	1	0	68.00	0	0	0	47	...	0	0	0	0	0	0	0	0	0	0
3	4616d365dd2853	a930a9c79cd721	f1f1f4ef412d7e	0	0	32.90	0	0	0	75	...	0	0	0	0	0	0	0	0	0	1
4	315c96c26c9aac	de04010b3458dd	6dc8ff871e21e6	0	0	100.00	0	0	0	45	...	0	0	0	0	0	0	0	1	0	0

5 rows × 51 columns

train['matchId'] = train['matchId'].astype('category')
train['groupId'] = train['groupId'].astype('category')

train.drop(columns=['Id'],inplace=True)

单排，双排，四排

solos=train[train['numGroups']>50]
duos=train[(train['numGroups']>25)&(train['numGroups']<=50)]
squads=train[train['numGroups']<=25]

len(solos)/len(train)

0.15947923787902407

len(duos)/len(train)

0.7412845072729117

f,ax =plt.subplots(figsize=(20,10))
sns.pointplot(x='kills',y='winPlacePerc',data=solos,color='black',alpha=0.8)
sns.pointplot(x='kills',y='winPlacePerc',data=duos,color='red',alpha=0.8)
sns.pointplot(x='kills',y='winPlacePerc',data=squads,color='blue',alpha=0.8)
plt.text(25,0.5,'Solos',color='red')
plt.grid()
plt.show()

corr

f,ax =plt.subplots(figsize=(15,15))
sns.heatmap(train.corr(),annot=True,linewidths=0.5,fmt='.1f',ax=ax)
plt.show()

k = 5
f,ax =plt.subplots(figsize=(12,12))
cols = train.corr().nlargest(k,'winPlacePerc')['winPlacePerc'].index
cm = np.corrcoef(train[cols].values.T)
sns.heatmap(cm,annot=True,linewidths=0.5,fmt='.1f',ax=ax,yticklabels=cols.values,xticklabels=cols.values)
plt.show()

建模

df_sample.head()

	assists	boosts	damageDealt	DBNOs	headshotKills	heals	killPlace	killPoints	kills	killStreaks	...	matchType_normal-duo	matchType_normal-duo-fpp	matchType_normal-solo	matchType_normal-solo-fpp	matchType_normal-squad	matchType_normal-squad-fpp	matchType_solo	matchType_solo-fpp	matchType_squad	matchType_squad-fpp
1980592	0	0	0.00	0	0	0	60	0	0	0	...	0	0	0	0	0	0	0	1	0	0
2448157	0	1	40.13	0	0	1	45	1024	0	0	...	0	0	0	0	0	0	0	0	1	0
2128667	0	0	0.00	0	0	0	70	0	0	0	...	0	0	0	0	0	0	0	0	1	0
3913225	0	0	0.00	0	0	0	79	0	0	0	...	0	0	0	0	0	0	0	0	0	0
2980818	0	0	46.92	0	0	1	52	0	0	0	...	0	0	0	0	0	0	0	0	0	1

5 rows × 48 columns

sample = 500000
df_sample = train.sample(sample)
df_sample.drop(columns = ['groupId','matchId'],inplace=True)

df = df_sample.drop(columns=['winPlacePerc'])
y = df_sample['winPlacePerc']

X_train,X_valid,y_train,y_valid = train_test_split(df,y,random_state=1)

def print_score(m):
    res= ['mae train',mean_absolute_error(m.predict(X_train),y_train),
         'mae val',mean_absolute_error(m.predict(X_valid),y_valid)]
    print (res)

from sklearn.metrics import mean_absolute_error
m1 = RandomForestRegressor(n_estimators=50,n_jobs=-1)
m1.fit(X_train,y_train)
print_score(m1)

['mae train', 0.021762396328252453, 'mae val', 0.058309784584455536]

m1.feature_importances_

array([1.35089375e-03, 4.87875324e-03, 3.32787366e-03, 3.68641782e-03,
       3.82363312e-04, 2.64745662e-03, 1.80408541e-01, 2.32876452e-03,
       2.69482315e-03, 2.92096788e-03, 5.21641404e-03, 8.68030154e-03,
       5.49998219e-03, 9.88430257e-03, 4.29501554e-03, 7.88414670e-04,
       1.91100321e-03, 4.66435583e-05, 6.18390314e-04, 2.47636965e-04,
       8.75905989e-05, 6.78730778e-01, 3.83837336e-03, 2.63527017e-03,
       1.83260527e-02, 9.11648154e-03, 3.63214124e-03, 1.17871960e-02,
       2.64370862e-02, 0.00000000e+00, 6.58062557e-04, 1.93873550e-05,
       1.03040286e-06, 2.40822626e-04, 4.06550376e-04, 4.99149059e-06,
       1.03070725e-05, 1.61284732e-06, 3.84765144e-05, 5.58591717e-06,
       1.44720204e-05, 1.58473274e-05, 1.79677780e-04, 1.27493547e-04,
       2.55436001e-04, 1.11125703e-03, 5.03059736e-04])

def rf_feat_importance(m,df):
    return pd.DataFrame({'cols':df.columns,'imp':m.feature_importances_}).sort_values('imp',ascending=False)

rf_feat_importance(m1,df)

	cols	imp
21	walkDistance	0.678731
6	killPlace	0.180409
28	totalDistance	0.026437
24	playersJoined	0.018326
27	matchDurationNorm	0.011787
13	numGroups	0.009884
25	killsNorm	0.009116
11	matchDuration	0.008680
12	maxPlace	0.005500
10	longestKill	0.005216
1	boosts	0.004879
14	rankPoints	0.004295
22	weaponsAcquired	0.003838
3	DBNOs	0.003686
26	damageDealtNorm	0.003632
2	damageDealt	0.003328
9	killStreaks	0.002921
8	kills	0.002695
5	heals	0.002647
23	winPoints	0.002635
7	killPoints	0.002329
16	rideDistance	0.001911
0	assists	0.001351
45	matchType_squad	0.001111
15	revives	0.000788
30	headshot_rate	0.000658
18	swimDistance	0.000618
46	matchType_squad-fpp	0.000503
34	matchType_duo-fpp	0.000407
4	headshotKills	0.000382
44	matchType_solo-fpp	0.000255
19	teamKills	0.000248
33	matchType_duo	0.000241
42	matchType_normal-squad-fpp	0.000180
43	matchType_solo	0.000127
20	vehicleDestroys	0.000088
17	roadKills	0.000047
38	matchType_normal-duo-fpp	0.000038
31	matchType_crashfpp	0.000019
41	matchType_normal-squad	0.000016
40	matchType_normal-solo-fpp	0.000014
36	matchType_flaretpp	0.000010
39	matchType_normal-solo	0.000006
35	matchType_flarefpp	0.000005
37	matchType_normal-duo	0.000002
32	matchType_crashtpp	0.000001
29	killWithoutMoving	0.000000

rf_feat_importance(m1,df)[:10].plot('cols','imp',figsize=(14,6),kind='barh')
plt.show()

fi=rf_feat_importance(m1,df)

to_keep = fi[fi.imp>0.02].cols
to_keep

21       walkDistance
6           killPlace
29      totalDistance
1              boosts
22    weaponsAcquired
5               heals
26    damageDealtNorm
2         damageDealt
16       rideDistance
10        longestKill
Name: cols, dtype: object

X_train,X_valid = X_train[to_keep],X_valid[to_keep]

m2 = RandomForestRegressor(n_estimators=50,n_jobs=-1)
m2.fit(X_train,y_train)
print_score(m2)

['mae train', 0.029341054777938774, 'mae val', 0.07770080750538866]