NBA运动员数据分析
背景信息
当前,篮球运动是最受欢迎的运动之一。在此万受瞩目的运动下,我打算针对篮球运动员个人的信息,技能水平等各项指标进行相关的分析与统计。例如,我们可能会关注如下的内容:
- 哪些球员从2014年到2019年近6年连续在榜?
- 对比各球员在2019年的各项数据如何?
- 詹姆斯-哈登随年份数据如何变化?
- 2019年球员数据分布如何?
- 篮球运动员的失误与上场时长有什么联系?
- 球员的各项数据之间的相关性如何?
- 哪些数据会对球员的得分有什么影响?
任务说明
概念
数据分析是指用适当的方法与工具,对收集来的大量数据进行分析,提取其中有意义的信息,从而形成有价值的结论的过程。
基本流程
在进行数据分析之前,需要清楚数据分析的基本流程。
- 明确需求与目的
- 分析篮球运动员,得出结论
- 数据收集
- 爬取新浪体育网站数据
- 数据预处理
- 特征筛选
- 降维
- 数据清洗
- 缺失值
- 异常值
- 重复值
- 特征筛选
- 数据分析
- 数据建模
- 数据可视化
- 编写报告总结
实验步骤
获取收集
导入相关库
import os
import requests
import re
import pandas as pd
from lxml import etree
import warnings
import matplotlib.pyplot as plt
import seaborn as sns
import pyecharts.options as opts
from pyecharts.globals import ThemeType
warnings.filterwarnings("ignore")
- 需要爬取的数据
- 爬虫思路分析:
- 1.需求获取新浪体育网站的球员信息
- 2.寻找网址(球员的不同赛季的信息在展示html页面找不到)
- 可判断该信息是一个Ajax请求
- 通过浏览器抓包,和多次尝试在Postman软件上发送request请求
- 寻得球员分页信息的request发送相应url的规律,并循环遍历
- 3.获取request返回的内容
- 4.使用正则表达式对request返回信息进行筛选
- 5.将筛选的数据以csv格式储存在本地文件中
# 爬虫程序
def spider_nba():
# 创建的目录
path = "./basketball_data"
if not os.path.exists(path):
os.mkdir(path)
for i in range(6):
url = 'http://slamdunk.sports.sina.com.cn/api?p=radar&callback=jQuery111306538669297726742_1571969723673&p=radar&s=leaders&a=players_top&season='+str(2019-i)+'&season_type=reg&item_type=average&item=points&order=1&_='+str(1571982115616+i)
response=requests.get(url)
# 采用utf-8解码
response.encoding='utf8'
# 读取reponse
data = response.text
data = re.findall('\{("rank.*?"personal_fouls":".*?")\}', data)
a_list = []
for item in data:
temp = item.split(",")
a_list.append(temp)
dic_a = dict()
for items in a_list:
for item in items:
key, value = item.split(":")
key = key[1:-1]
if key == '':
continue
if value[1:3] == "\\u":
value = "u" + value
value = eval(value)
else:
value = value[1:-1]
if re.match("^\d+$", value):
value = int(value)
elif re.match("^\d*\.\d+$", value):
value = float(value)
if key not in dic_a.keys():
dic_a[key] = [value]
else:
dic_a[key].append(value)
# 用数据框接收
df = pd.DataFrame(dic_a)
# 写入文件保存
df.to_csv('./basketball_data/player_'+str(2019-i)+'.csv')
# 判断./basketball_data/路径下是否存在文件,如果不存在,就执行爬虫程序
if not os.path.exists('./basketball_data/'):
spider_nba()
else:
pass
# 查看写入数据结果
# os.getcwd()
# os.chdir(path)
# 修改当前路径
os.chdir(path='./basketball_data/')
print(os.getcwd())
os.listdir()
['player_2014.csv', 'player_2015.csv', 'player_2016.csv', 'player_2017.csv', 'player_2018.csv', 'player_2019.csv']
# 2014年到2019年的球员信息存储到字典中
year_df = dict()
for i in range(6):
df = pd.read_csv('player_'+str(2019-i)+'.csv')
try:
df = df.drop("Unnamed: 0",axis=1)
except:
pass
year_df[2019-i] = df
for i in range(2014, 2020):
print("%d年数据形状: %d行 %d列" % (i, year_df[i].shape[0], year_df[i].shape[1]))
2014年数据形状: 25行 29列 2015年数据形状: 25行 29列 2016年数据形状: 25行 29列 2017年数据形状: 25行 29列 2018年数据形状: 25行 29列 2019年数据形状: 27行 29列
# 查看列名
year_df[2019].columns
Index(['rank', 'score', 'pid', 'first_name', 'last_name', 'tid', 'team_name', 'games_played', 'games_started', 'minutes', 'points', 'field_goals_made', 'field_goals_att', 'field_goals_pct', 'three_points_made', 'three_points_att', 'three_points_pct', 'free_throws_made', 'free_throws_att', 'free_throws_pct', 'offensive_rebounds', 'defensive_rebounds', 'rebounds', 'assists', 'turnovers', 'assists_turnover_ratio', 'steals', 'blocks', 'personal_fouls'], dtype='object')
# 查看前三行
year_df[2019].head(3)
rank | score | pid | first_name | last_name | tid | team_name | games_played | games_started | minutes | ... | free_throws_pct | offensive_rebounds | defensive_rebounds | rebounds | assists | turnovers | assists_turnover_ratio | steals | blocks | personal_fouls | |
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
0 | 1 | 38.5 | cf418e0c-de9d-438f-a1ac-3be539a56c42 | 特雷 | 杨 | 583ecb8f-fb46-11e1-82cb-f4ce4684ea4c | 老鹰 | 2 | 2 | 36.5 | ... | 0.750 | 0.0 | 7.0 | 7.0 | 9.0 | 5.5 | 1.6 | 1.5 | 0.00 | 1.0 |
1 | 2 | 37.7 | dd146010-902b-4ad7-b98c-650d0363a2f0 | 凯里 | 欧文 | 583ec9d6-fb46-11e1-82cb-f4ce4684ea4c | 篮网 | 3 | 3 | 34.7 | ... | 0.931 | 1.3 | 4.3 | 5.7 | 6.3 | 2.0 | 3.2 | 1.7 | 0.67 | 3.3 |
2 | 3 | 32.0 | ab532a66-9314-4d57-ade7-bb54a70c65ad | 卡尔-安东尼 | 唐斯 | 583eca2f-fb46-11e1-82cb-f4ce4684ea4c | 森林狼 | 3 | 3 | 33.7 | ... | 0.630 | 2.3 | 11.0 | 13.3 | 5.0 | 2.3 | 2.1 | 2.7 | 2.00 | 3.0 |
3 rows × 29 columns
数据预处理
特征筛选
- 数据集中的列,并非都是我们分析所需要的
- 可以有选择性的进行加载,只加载我们需要的信息列
for i in range(2014, 2020):
year_df[i]
# 删除以下列
del_name = ['pid','tid','games_played','games_started','points']
year_df[i] = year_df[i].drop(del_name,axis=1)
# 连接first_name和last_name
year_df[i]['player_name'] = year_df[i]['first_name']+"-"+year_df[i]['last_name']
player_name = year_df[i].player_name
year_df[i] = year_df[i].drop(['first_name','last_name'],axis=1)
year_df[i] = year_df[i].drop('player_name',axis=1)
# 将player_name插入到第二列
year_df[i].insert(1,'player_name',player_name)
team_name = df.team_name
year_df[i] = year_df[i].drop('team_name',axis=1)
# 将team_name插入到第三列
year_df[i].insert(2,'team_name',team_name)
year_df[2019].columns
Index(['rank', 'player_name', 'team_name', 'score', 'minutes', 'field_goals_made', 'field_goals_att', 'field_goals_pct', 'three_points_made', 'three_points_att', 'three_points_pct', 'free_throws_made', 'free_throws_att', 'free_throws_pct', 'offensive_rebounds', 'defensive_rebounds', 'rebounds', 'assists', 'turnovers', 'assists_turnover_ratio', 'steals', 'blocks', 'personal_fouls'], dtype='object')
# 查看前三行
year_df[2019].head(3)
rank | player_name | team_name | score | minutes | field_goals_made | field_goals_att | field_goals_pct | three_points_made | three_points_att | ... | free_throws_pct | offensive_rebounds | defensive_rebounds | rebounds | assists | turnovers | assists_turnover_ratio | steals | blocks | personal_fouls | |
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
0 | 1 | 特雷-杨 | 雷霆 | 38.5 | 36.5 | 13.5 | 23.0 | 0.587 | 5.5 | 10.0 | ... | 0.750 | 0.0 | 7.0 | 7.0 | 9.0 | 5.5 | 1.6 | 1.5 | 0.00 | 1.0 |
1 | 2 | 凯里-欧文 | 火箭 | 37.7 | 34.7 | 12.0 | 26.3 | 0.456 | 4.7 | 11.3 | ... | 0.931 | 1.3 | 4.3 | 5.7 | 6.3 | 2.0 | 3.2 | 1.7 | 0.67 | 3.3 |
2 | 3 | 卡尔-安东尼-唐斯 | 雷霆 | 32.0 | 33.7 | 10.7 | 20.3 | 0.525 | 5.0 | 9.7 | ... | 0.630 | 2.3 | 11.0 | 13.3 | 5.0 | 2.3 | 2.1 | 2.7 | 2.00 | 3.0 |
3 rows × 23 columns
数据清洗
缺失值处理
- 通过info查看数据信息。
- 可以通过isnull与sum结合,查看缺失值情况。
# info方法可以显示每列名称,非空值数量,每列的数据类型,内存占用等信息。
# data.info()
for i in range(2014, 2019):
print("============"+str(i)+"年============")
print(year_df[i].isnull().sum(axis=0))
==2014年=
rank 0
player_name 0
team_name 0
score 0
minutes 0
field_goals_made 0
field_goals_att 0
field_goals_pct 0
three_points_made 0
three_points_att 0
three_points_pct 0
free_throws_made 0
free_throws_att 0
free_throws_pct 0
offensive_rebounds 0
defensive_rebounds 0
rebounds 0
assists 0
turnovers 0
assists_turnover_ratio 0
steals 0
blocks 0
personal_fouls 0
dtype: int64
2015年
rank 0
player_name 0
team_name 0
score 0
minutes 0
field_goals_made 0
field_goals_att 0
field_goals_pct 0
three_points_made 0
three_points_att 0
three_points_pct 0
free_throws_made 0
free_throws_att 0
free_throws_pct 0
offensive_rebounds 0
defensive_rebounds 0
rebounds 0
assists 0
turnovers 0
assists_turnover_ratio 0
steals 0
blocks 0
personal_fouls 0
dtype: int64
2016年
rank 0
player_name 0
team_name 0
score 0
minutes 0
field_goals_made 0
field_goals_att 0
field_goals_pct 0
three_points_made 0
three_points_att 0
three_points_pct 0
free_throws_made 0
free_throws_att 0
free_throws_pct 0
offensive_rebounds 0
defensive_rebounds 0
rebounds 0
assists 0
turnovers 0
assists_turnover_ratio 0
steals 0
blocks 0
personal_fouls 0
dtype: int64
2017年
rank 0
player_name 0
team_name 0
score 0
minutes 0
field_goals_made 0
field_goals_att 0
field_goals_pct 0
three_points_made 0
three_points_att 0
three_points_pct 0
free_throws_made 0
free_throws_att 0
free_throws_pct 0
offensive_rebounds 0
defensive_rebounds 0
rebounds 0
assists 0
turnovers 0
assists_turnover_ratio 0
steals 0
blocks 0
personal_fouls 0
dtype: int64
2018年
rank 0
player_name 0
team_name 0
score 0
minutes 0
field_goals_made 0
field_goals_att 0
field_goals_pct 0
three_points_made 0
three_points_att 0
three_points_pct 0
free_throws_made 0
free_throws_att 0
free_throws_pct 0
offensive_rebounds 0
defensive_rebounds 0
rebounds 0
assists 0
turnovers 0
assists_turnover_ratio 0
steals 0
blocks 0
personal_fouls 0
dtype: int64
# 删除所有含有空值的行。就地修改。
# year_df[2019].dropna(axis=0, inplace=True)
# year_df[2019].isnull().sum()
异常值处理
- 通过describe查看数值信息。
- 可配合箱线图辅助。
- 异常值可以删除,视为缺失值,或者不处理。
year_df[2019].describe()
rank | score | minutes | field_goals_made | field_goals_att | field_goals_pct | three_points_made | three_points_att | three_points_pct | free_throws_made | ... | free_throws_pct | offensive_rebounds | defensive_rebounds | rebounds | assists | turnovers | assists_turnover_ratio | steals | blocks | personal_fouls | |
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
count | 27.000000 | 27.000000 | 27.000000 | 27.000000 | 27.000000 | 27.000000 | 27.000000 | 27.000000 | 27.000000 | 27.000000 | ... | 27.000000 | 27.000000 | 27.000000 | 27.000000 | 27.000000 | 27.000000 | 27.000000 | 27.000000 | 27.000000 | 27.000000 |
mean | 13.481481 | 26.211111 | 33.692593 | 9.192593 | 18.529630 | 0.501481 | 2.290000 | 6.070370 | 0.386704 | 5.551852 | ... | 0.825852 | 1.235556 | 6.200000 | 7.437037 | 5.611111 | 3.285185 | 1.932222 | 1.277407 | 0.865185 | 3.007407 |
std | 7.412932 | 4.497122 | 2.674944 | 1.712606 | 2.953201 | 0.086392 | 1.332554 | 3.169288 | 0.156651 | 2.243344 | ... | 0.126343 | 1.066000 | 3.274376 | 3.818694 | 2.856078 | 1.438670 | 1.286654 | 0.695500 | 0.909105 | 1.273676 |
min | 1.000000 | 22.000000 | 26.300000 | 5.000000 | 13.000000 | 0.238000 | 0.000000 | 0.330000 | 0.000000 | 1.700000 | ... | 0.550000 | 0.000000 | 2.000000 | 2.300000 | 1.700000 | 1.000000 | 0.570000 | 0.330000 | 0.000000 | 1.000000 |
25% | 7.500000 | 23.150000 | 32.250000 | 8.300000 | 16.500000 | 0.458000 | 1.400000 | 4.350000 | 0.321500 | 4.150000 | ... | 0.756000 | 0.330000 | 3.850000 | 4.500000 | 3.300000 | 2.300000 | 1.150000 | 0.585000 | 0.330000 | 2.150000 |
50% | 14.000000 | 25.000000 | 34.000000 | 9.000000 | 19.000000 | 0.508000 | 2.000000 | 5.500000 | 0.385000 | 5.500000 | ... | 0.833000 | 1.000000 | 5.000000 | 6.000000 | 5.000000 | 3.300000 | 1.600000 | 1.300000 | 0.500000 | 3.000000 |
75% | 20.000000 | 28.000000 | 35.650000 | 10.150000 | 20.300000 | 0.555500 | 3.000000 | 8.350000 | 0.500000 | 6.500000 | ... | 0.929000 | 1.700000 | 8.300000 | 10.150000 | 8.150000 | 4.000000 | 2.400000 | 1.700000 | 1.500000 | 3.850000 |
max | 23.000000 | 38.500000 | 37.300000 | 13.500000 | 26.300000 | 0.646000 | 5.500000 | 13.000000 | 0.750000 | 12.500000 | ... | 1.000000 | 3.700000 | 13.500000 | 15.700000 | 10.500000 | 7.500000 | 7.000000 | 2.700000 | 3.300000 | 6.000000 |
8 rows × 21 columns
# 箱型图
plt.figure(figsize=(15, 8))
df = year_df[2019].iloc[:, 3:].copy()
col_name_fe = []
col_name_yi = dict()
i = 0
for item in df.columns.values:
temp = (item[0] + item[1] + item[-2]).upper()
col_name_fe.append(temp)
col_name_yi[temp.upper()] = item
i += 1
df.columns = col_name_fe
# whitegrid,darkgrid
sns.set_style("whitegrid")
sns.boxplot(data=df[list(df.columns)])
print(col_name_yi)
#小于q1 - 1.5IQR 大于q3 + 1.5IQR
{‘SCR’: ‘score’, ‘MIE’: ‘minutes’, ‘FID’: ‘field_goals_made’, ‘FIT’: ‘field_goals_att’, ‘FIC’: ‘field_goals_pct’, ‘THD’: ‘three_points_made’, ‘THT’: ‘three_points_att’, ‘THC’: ‘three_points_pct’, ‘FRD’: ‘free_throws_made’, ‘FRT’: ‘free_throws_att’, ‘FRC’: ‘free_throws_pct’, ‘OFD’: ‘offensive_rebounds’, ‘DED’: ‘defensive_rebounds’, ‘RED’: ‘rebounds’, ‘AST’: ‘assists’, ‘TUR’: ‘turnovers’, ‘ASI’: ‘assists_turnover_ratio’, ‘STL’: ‘steals’, ‘BLK’: ‘blocks’, ‘PEL’: ‘personal_fouls’}
重复值处理
- 使用duplicate检查重复值。可配合keep参数进行调整。
- 使用drop_duplicate删除重复值。
year_df[2019].duplicated().sum()
# data.drop_duplicates(inplace=True)
0
列名英文名词解释
# 中文列命名
cn_name = [
'排名','球员姓名','球队名称','得分','上场时间','投篮命中数',
'投篮数','投篮命中率','三分命中数','三分球数','三分命中率',
'罚球命中数','罚球数','罚球命中率','进攻篮板','防守篮板',
'总篮板','助攻','失误','助攻率','抢断','盖帽','犯规'
]
# 英文指标解释
en_name = year_df[2019].columns.values.tolist()
col_en_cn_name = [en_name, cn_name]
temp_df = pd.DataFrame(col_en_cn_name).T
temp_df.columns=["英文", "中文"]
temp_df
英文 | 中文 | |
---|---|---|
0 | rank | 排名 |
1 | player_name | 球员姓名 |
2 | team_name | 球队名称 |
3 | score | 得分 |
4 | minutes | 上场时间 |
5 | field_goals_made | 投篮命中数 |
6 | field_goals_att | 投篮数 |
7 | field_goals_pct | 投篮命中率 |
8 | three_points_made | 三分命中数 |
9 | three_points_att | 三分球数 |
10 | three_points_pct | 三分命中率 |
11 | free_throws_made | 罚球命中数 |
12 | free_throws_att | 罚球数 |
13 | free_throws_pct | 罚球命中率 |
14 | offensive_rebounds | 进攻篮板 |
15 | defensive_rebounds | 防守篮板 |
16 | rebounds | 总篮板 |
17 | assists | 助攻 |
18 | turnovers | 失误 |
19 | assists_turnover_ratio | 助攻率 |
20 | steals | 抢断 |
21 | blocks | 盖帽 |
22 | personal_fouls | 犯规 |
数据分析
哪些球员近6年连续在榜
# 哪些球员从2014年到2019年一直在榜
temp_list = []
temp_set = set()
for i in range(2014, 2020):
temp_list.append(set(year_df[i]['player_name']))
temp_set = temp_list[0]&temp_list[1]&temp_list[2]&temp_list[3]&temp_list[4]&temp_list[5]
temp_set
{'拉塞尔-威斯布鲁克', '詹姆斯-哈登', '达米安-利拉德'}
19年球员各项数据对比
from pyecharts.charts import Bar
# 2019年各球员各项数据对比
def bar_datazoom_slider() -> Bar:
c = (
Bar(init_opts=opts.InitOpts(theme=ThemeType.VINTAGE))
.add_xaxis(cn_name[3:])
.add_yaxis("凯里-欧文", year_df[2019].iloc[0, 3:].tolist())
.add_yaxis("特雷-杨", year_df[2019].iloc[1, 3:].tolist())
.add_yaxis("帕斯卡尔-西亚卡姆", year_df[2019].iloc[4, 3:].tolist())
.add_yaxis("丹尼-格林", year_df[2019].iloc[10, 3:].tolist())
.add_yaxis("考瓦伊-莱昂纳德", year_df[2019].iloc[18, 3:].tolist())
.add_yaxis("安东尼-戴维斯", year_df[2019].iloc[22, 3:].tolist())
.add_yaxis("拉塞尔-威斯布鲁克", year_df[2019].iloc[24, 3:].tolist())
.set_global_opts(
title_opts=opts.TitleOpts(title='''
2019年各球员各项数据对比'''),
datazoom_opts=opts.DataZoomOpts(),
)
)
return c
bar_datazoom_slider().render_notebook()
19年球员数据分布箱型图
# 查看2019NBA球员数据分布箱型图
from pyecharts.charts import Boxplot
def boxpolt_base() -> Boxplot:
v1 = [year_df[2019]['score'].tolist()]
v2 = [year_df[2019]['minutes'].tolist()]
v3 = [year_df[2019]['field_goals_att'].tolist()]
v4 = [year_df[2019]['three_points_att'].tolist()]
v5 = [year_df[2019]['rebounds'].tolist()]
v6 = [year_df[2019]['assists'].tolist()]
c = Boxplot(init_opts=opts.InitOpts(theme=ThemeType.VINTAGE))
c.add_xaxis([]).add_yaxis(
"score", c.prepare_data(v1)).add_yaxis(
"minutes", c.prepare_data(v2)).add_yaxis(
"field_goals_att", c.prepare_data(v3)).add_yaxis(
"three_points_att", c.prepare_data(v4)).add_yaxis(
"rebounds", c.prepare_data(v5)).add_yaxis(
"assists", c.prepare_data(v6)
).set_global_opts(title_opts=opts.TitleOpts(title=
'''2019年
NBA球员数据箱型图'''))
return c
boxpolt_base().render_notebook()
詹姆斯-哈登随年份数据变化折线图
# 詹姆斯-哈登随年份数据变化折线图
year_name = list(range(2014, 2020))
name_year_dic = dict()
# for name in playe_name:
for year in year_name:
temp = year_df[year][year_df[year]['player_name'].isin(['拉塞尔-威斯布鲁克'])]
year_list = temp.values.tolist()
year_list = [item for items in year_list for item in items]
name_year_dic[year] = year_list
name_year_df = pd.DataFrame(name_year_dic)
name_year_df.index = cn_name
name_year_df.head()
s1 = list(name_year_df.loc['得分', :])
s2 = list(name_year_df.loc['上场时间', :])
s3 = list(name_year_df.loc['投篮命中率', :])
s4 = list(name_year_df.loc['三分命中率', :])
s5 = list(name_year_df.loc['罚球命中率', :])
s6 = list(name_year_df.loc['三分命中数', :])
s7 = list(name_year_df.loc['进攻篮板', :])
s8 = list(name_year_df.loc['失误', :])
s9 = list(name_year_df.loc['助攻率', :])
s10 = list(name_year_df.loc['抢断', :])
s11 = list(name_year_df.loc['盖帽', :])
s12 = list(name_year_df.loc['犯规', :])
from pyecharts.charts import Line
def line_base() -> Line:
c = (
Line(init_opts=opts.InitOpts(theme=ThemeType.VINTAGE))
.add_xaxis(list(map(str,year_name)))
.add_yaxis("得分", s1)
.add_yaxis("上场时间", s2)
.add_yaxis("投篮命中率", s3)
.add_yaxis("三分命中率", s4)
.add_yaxis("罚球命中率", s5)
.add_yaxis("进攻篮板", s6)
.add_yaxis("失误", s7)
.add_yaxis("助攻率", s8)
.add_yaxis("抢断", s9)
.add_yaxis("盖帽", s10)
.add_yaxis("犯规", s11)
.set_global_opts(title_opts=opts.TitleOpts(title='''
詹姆斯-哈登随年份数据变化折线图'''))
)
return c
line_base().render_notebook()
篮球运动员的失误与上场时长有什么联系?
sns.scatterplot(x="minutes", y="rebounds", data=year_df[2019])
# 两变量的相关系数
year_df[2019]["minutes"].corr(year_df[2019]["rebounds"])
0.13196323727585887
各数据的相关性图
# data.corr()
# 相关性图
plt.figure(figsize=(25, 12))
sns.heatmap(year_df[2019].corr(), annot=True, fmt=".2f", cmap=plt.cm.Greens)
# plt.savefig("../corr.png", dpi=100, bbox_inches="tight")
画决策树分析影响得分的因素
# 增加年份一列,并合并2014-2019年所有球员信息
year_df_new = pd.DataFrame()
for year in year_name:
if year == 2014:
year_df_temp = year_df[year]
year_df_temp['year'] = year
year_df_new = year_df_temp
else:
year_df_temp = year_df[year]
year_df_temp['year'] = year
year_df_new = year_df_new.append(year_df_temp)
year_df_new.shape
(152, 24)
set(year_df_new['year'])
{2014, 2015, 2016, 2017, 2018, 2019}
year_df_new.columns
Index([‘rank’, ‘player_name’, ‘team_name’, ‘score’, ‘minutes’,
‘field_goals_made’, ‘field_goals_att’, ‘field_goals_pct’,
‘three_points_made’, ‘three_points_att’, ‘three_points_pct’,
‘free_throws_made’, ‘free_throws_att’, ‘free_throws_pct’,
‘offensive_rebounds’, ‘defensive_rebounds’, ‘rebounds’, ‘assists’,
‘turnovers’, ‘assists_turnover_ratio’, ‘steals’, ‘blocks’,
‘personal_fouls’, ‘year’],
dtype=‘object’)
# 删除不需要的列
year_df_new = year_df_new.drop(['rank', 'player_name', 'team_name'],axis=1)
year_df_new.columns
Index([‘score’, ‘minutes’, ‘field_goals_made’, ‘field_goals_att’,
‘field_goals_pct’, ‘three_points_made’, ‘three_points_att’,
‘three_points_pct’, ‘free_throws_made’, ‘free_throws_att’,
‘free_throws_pct’, ‘offensive_rebounds’, ‘defensive_rebounds’,
‘rebounds’, ‘assists’, ‘turnovers’, ‘assists_turnover_ratio’, ‘steals’,
‘blocks’, ‘personal_fouls’, ‘year’],
dtype=‘object’)
# 保存year_df_new数据
year_df_new.to_csv('../year_df_new.csv')
# 分析影响得分的因素
dataSet = pd.read_csv('../year_df_new.csv', dtype = {'year' : float})
dataSet = dataSet.drop('Unnamed: 0',axis=1)
dataSet.dtypes
score float64
minutes float64
field_goals_made float64
field_goals_att float64
field_goals_pct float64
three_points_made float64
three_points_att float64
three_points_pct float64
free_throws_made float64
free_throws_att float64
free_throws_pct float64
offensive_rebounds float64
defensive_rebounds float64
rebounds float64
assists float64
turnovers float64
assists_turnover_ratio float64
steals float64
blocks float64
personal_fouls float64
year float64
dtype: object
#自定义区间并进行分割
qujian=[0,25,100]
pd.cut(dataSet.score,qujian)
#起别名
dataSet['score'] = pd.cut(dataSet.score,qujian,labels=[1,2])
# dataSet['score']
dataSet.head()
score | minutes | field_goals_made | field_goals_att | field_goals_pct | three_points_made | three_points_att | three_points_pct | free_throws_made | free_throws_att | ... | offensive_rebounds | defensive_rebounds | rebounds | assists | turnovers | assists_turnover_ratio | steals | blocks | personal_fouls | year | |
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
0 | 2 | 34.4 | 9.4 | 22.0 | 0.426 | 1.30 | 4.30 | 0.299 | 8.2 | 9.8 | ... | 1.90 | 5.4 | 7.3 | 8.6 | 4.4 | 2.0 | 2.10 | 0.21 | 2.8 | 2014.0 |
1 | 2 | 36.8 | 8.0 | 18.2 | 0.440 | 2.60 | 6.90 | 0.375 | 8.8 | 10.2 | ... | 0.93 | 4.7 | 5.7 | 7.0 | 4.0 | 1.8 | 1.90 | 0.74 | 2.6 | 2014.0 |
2 | 2 | 33.8 | 8.8 | 17.3 | 0.510 | 2.40 | 5.90 | 0.403 | 5.4 | 6.3 | ... | 0.59 | 6.0 | 6.6 | 4.1 | 2.7 | 1.5 | 0.89 | 0.93 | 1.5 | 2014.0 |
3 | 2 | 36.1 | 9.0 | 18.5 | 0.488 | 1.70 | 4.90 | 0.354 | 5.4 | 7.7 | ... | 0.74 | 5.3 | 6.0 | 7.4 | 3.9 | 1.9 | 1.60 | 0.71 | 2.0 | 2014.0 |
4 | 1 | 36.1 | 9.4 | 17.6 | 0.535 | 0.01 | 0.18 | 0.083 | 5.5 | 6.8 | ... | 2.50 | 7.7 | 10.2 | 2.2 | 1.4 | 1.6 | 1.50 | 2.90 | 2.1 | 2014.0 |
5 rows × 21 columns
from itertools import product
import numpy as np
import matplotlib.pyplot as plt
from sklearn import datasets
from sklearn.tree import DecisionTreeClassifier
from IPython.display import Image
from sklearn import tree
import pydotplus
import os
os.environ["PATH"] += os.pathsep + 'C:/Program Files (x86)/Graphviz2.38/bin'
X = dataSet.iloc[:, 1:]
y = dataSet.iloc[:, 0]
# 训练模型,限制树的最大深度4
clf = DecisionTreeClassifier(max_depth=4)
#拟合模型
clf.fit(X, y)
dot_data = tree.export_graphviz(clf, out_file=None,
feature_names=dataSet.columns.values[1:],
class_names=['low', 'high'],
filled=True, rounded=True,
special_characters=True)
graph = pydotplus.graph_from_dot_data(dot_data)
# 使用ipython的终端jupyter notebook显示。
Image(graph.create_png())
# 如果没有ipython的jupyter notebook,可以把此图写到pdf文件里,在pdf文件里查看。
# graph.write_pdf("iris.pdf")
总结
- 拉塞尔-威斯布鲁克、詹姆斯-哈登、达米安-利拉德这三人连续6年从2014年到2020年都在榜上,可以说明这三人在当下球员竞技状态非常稳定
- 对比2019年各球员各项数据,可知球员与球员之间相同数据项的差异
- 针对詹姆斯-哈登,分析其六年的各项数据的变化,2016年其各项正数据普偏高,而负数据普偏低,所有该年是詹姆斯-哈登的巅峰期。从宏观上看,从2014年打篮球能力从2014年到2016年逐步提高,到2016年达到巅峰,之后下降,最后趋向与平稳
- 可根据箱型图可视化所有球员各项数据分布
- 篮球运动员的失误与上场时长成正相关
- 分析球员的各项数据之间的相关性,可知得分与投球的数据关系非常大,其次是投球命中率
- 根据决策树逐步分析各项因素决定得分情况,由此可以对球员各项训练计划表进行优化