科比生涯数据分析——利用随机森林进行分类

1.数据读取与介绍

• 导入相关库及模块

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from sklearn.ensemble import RandomForestClassifier
from sklearn.model_selection import KFold
           

• 利用pandas进行数据读取，通过info()函数了解该数据的大致信息

file_name='data.csv'
data=pd.read_csv(file_name)
print('****该数据的大致信息如下****')
print(data.info())
           

• 打印前五行数据

• 获取该数据的行数与列数

输出为：该数据共有30697条记录，25个特征项

通过以上结果显示，我们得到：该数据在shot_made_flag字段上缺失值较多，且该字段为标签项，0表示未射入球门，1表示摄入球门，所以需删去shot_made_flag项为NaN的记录行。

data=data[data['shot_made_flag'].notnull()]
data.info()
           

2.特征数据可视化展示

• 将射球时相对于球门的位置(loc_x,loc_y)，(lat,lon)在图形中展示出来

#设置画布大小
plt.figure(figsize=(12,12))
#画第一个子图
plt.subplot(121)
plt.title('the location of the shot')
plt.xlabel('loc_x')
plt.ylabel('loc_y')
plt.scatter(data['loc_x'], data['loc_y'], color='g', alpha = 0.02)
#画第二个子图
plt.subplot(122)
plt.title('the site of the shot')
plt.xlabel('longitude')#经度
plt.ylabel('latitude')#纬度
plt.scatter(data['lon'], data['lat'], color='r', alpha = 0.02)
           

3.数据预处理

上面可视化图形显示，科比的射门位置大致呈半圆形，因此构建两个新的字段值dist和angle，其中dist=sqtr(x^2 +y^2)，angle为夹角。

data['dist'] = np.sqrt(data['loc_x']**2 + data['loc_y']**2)

loc_x_zero = data['loc_x'] == 0
#print (loc_x_zero)
data['angle'] = np.array([0]*len(data))
data['angle'][~loc_x_zero] = np.arctan(data['loc_y'][~loc_x_zero] / data['loc_x'][~loc_x_zero])
data['angle'][loc_x_zero] = np.pi / 2 
           

• 构建新的字段remaining_time

• 打印字段action_type、combined_shot_type、shot_type和shot_type

print(data.action_type.unique())
print(data.combined_shot_type.unique())
print(data.shot_type.unique())
print(data.shot_type.value_counts())
           

打印字段season

输出：array([‘2000-01’, ‘2001-02’, ‘2002-03’, ‘2003-04’, ‘2004-05’, ‘2005-06’,

‘2006-07’, ‘2007-08’, ‘2008-09’, ‘2009-10’, ‘2010-11’, ‘2011-12’,

‘2012-13’, ‘2013-14’, ‘2014-15’, ‘2015-16’, ‘1996-97’, ‘1997-98’,

‘1998-99’, ‘1999-00’], dtype=object)

• 构建新列

data['season'] = data['season'].apply(lambda x: int(x.split('-')[1]) )
data['season'].unique()
           

• 可视化distance与dist之间的关系

plt.figure(figsize=(5,5))
plt.scatter(data['dist'], data['shot_distance'], color='blue')
plt.title('dist and shot_distance')
plt.xlabel('dist')
plt.ylabel('shot_distance')
           

• 删除多余字段

drops = ['shot_id', 'team_id', 'team_name', 'shot_zone_area', 'shot_zone_range', 'shot_zone_basic', \
         'matchup', 'lon', 'lat', 'seconds_remaining', 'minutes_remaining', \
         'shot_distance', 'loc_x', 'loc_y', 'game_event_id', 'game_id', 'game_date']
for drop in drops:
    data = data.drop(drop, 1)
           

• 将分类字段转化为数值型

categorical_vars = ['action_type', 'combined_shot_type', 'shot_type', 'opponent', 'period', 'season']
for var in categorical_vars:
    data = pd.concat([data, pd.get_dummies(data[var], prefix=var)], 1)
    data = data.drop(var, 1)
           

4.使用scikit-learn建立模型

• 构造训练集

train_kobe = data.copy()
train_kobe = train_kobe.drop(axis=1, columns='shot_made_flag')
train_label = data['shot_made_flag']
           

• 导入相关库及模块

from sklearn.ensemble import RandomForestRegressor
from sklearn.metrics import confusion_matrix,log_loss
import time
from sklearn.ensemble import RandomForestClassifier
from sklearn.model_selection import KFold
           

• 机器学习步骤如下：

  在这里面，我们采用随机森林集成算法，对科比是否进球进行合理分类，代码主要实现目标为：

• 寻求随机森林中树的最优构建数量
• 寻求树的最优深度，防止过度拟合现象发生

# find the best n_estimators for RandomForestClassifier
print('Finding best n_estimators for RandomForestClassifier...')
min_score = 100000
best_n = 0
scores_n = []
range_n = np.logspace(0,2,num=3).astype(int)
for n in range_n:
    print("the number of trees : {0}".format(n))
    t1 = time.time()    
    rfc_score = 0.
    rfc = RandomForestClassifier(n_estimators=n)
    for train_k, test_k in KFold(len(train_kobe), n_folds=10, shuffle=True):
        rfc.fit(train_kobe.iloc[train_k], train_label.iloc[train_k])
        #rfc_score += rfc.score(train.iloc[test_k], train_y.iloc[test_k])/10
        pred = rfc.predict(train_kobe.iloc[test_k])
        rfc_score += log_loss(train_label.iloc[test_k], pred) / 10
    scores_n.append(rfc_score)
    if rfc_score < min_score:
        min_score = rfc_score
        best_n = n
        
    t2 = time.time()
    print('Done processing {0} trees ({1:.3f}sec)'.format(n, t2-t1))
print(best_n, min_score)

# find best max_depth for RandomForestClassifier
print('Finding best max_depth for RandomForestClassifier...')
min_score = 100000
best_m = 0
scores_m = []
range_m = np.logspace(0,2,num=3).astype(int)
for m in range_m:
    print("the max depth : {0}".format(m))
    t1 = time.time()   
    rfc_score = 0.
    rfc = RandomForestClassifier(max_depth=m, n_estimators=best_n)
    for train_k, test_k in KFold(len(train_kobe), n_folds=10, shuffle=True):
        rfc.fit(train_kobe.iloc[train_k], train_label.iloc[train_k])
        #rfc_score += rfc.score(train.iloc[test_k], train_y.iloc[test_k])/10
        pred = rfc.predict(train_kobe.iloc[test_k])
        rfc_score += log_loss(train_label.iloc[test_k], pred) / 10
    scores_m.append(rfc_score)
    if rfc_score < min_score:
        min_score = rfc_score
        best_m = m
    
    t2 = time.time()
    print('Done processing {0} trees ({1:.3f}sec)'.format(m, t2-t1))
print(best_m, min_score)
           

输出结果如下：

结果显示：最优树数量为100，树最大深度为10

• 可视化在不同数量树以及树深度下随机森林的信息熵（香农熵），信息熵越大，表明此时信息不确定性越大，即预测的准确性越低