importpandas as pdimportmathimportcsvimportrandomimportnumpy as npfrom sklearn importlinear_modelfrom sklearn.model_selection importcross_val_score

base_elo= 1600team_elos={}

team_stats={}

X=[]

y=[]#初始化数据，从T，O，M表格中读取数据，取出一些无关数据并将这三个表格通过team树形列进行连接：#根据每个队伍的Miscellaneous Opponent，Team统计数据csv文件进行初始化

definitialize_data(Mstat,Ostat,Tstat):

new_Mstat= Mstat.drop(['Rk','Arena'],axis=1)

new_Ostat= Ostat.drop(['Rk',"G",'MP'],axis=1)

new_Tstat= Tstat.drop(['Rk',"G",'MP'],axis=1)

team_stats1= pd.merge(new_Mstat,new_Ostat,how='left',on='Team')

team_stats1= pd.merge(team_stats1,new_Tstat,how='left',on='Team')return team_stats1.set_index('Team',inplace=False,drop=True)defget_elo(team):try:returnteam_elos[team]except:

team_elos[team]=base_eloreturnteam_elos[team]defcalc_elo(win_team,lose_team):

winner_rank=get_elo(win_team)

loser_rank=get_elo(lose_team)#根据Logistic Distribution计算 PK 双方(A和B)对各自的胜率期望值计算公式

rank_diff = winner_rank -loser_rank

exp= (rank_diff *-1)/400odds= 1/(1+math.pow(10,exp))#根据rank界别修改k值

if winner_rank < 2100:

k= 32

elif winner_rank >=2100 and winner_rank <2400:

k= 24

else:

k=16

#更新rank数值

new_winner_rank = round(winner_rank+(k*(1-odds)))

new_loser_rank= round(loser_rank+(k*(0-odds)))returnnew_winner_rank,new_loser_rank#基于统计好的数据，给每只队伍的eloscore计算结果，建立对应15-16年数据集，我们认为主场作战的队伍更有优势，因此会给主场队伍加上100分

defbuild_dataSet(all_data):print("Building data set..")

X=[]

skip=0for index,row inall_data.iterrows():

Wteam= row['WTeam']

Lteam= row['LTeam']#获取最初的elo或者每个队伍最初的elo值

team1_elo =get_elo(Wteam)

team2_elo=get_elo(Lteam)#给主场比赛队伍加上100的elo值

if row['WLoc'] == 'H':

team1_elo+= 100

else:

team2_elo+= 100

#把elo当成评价每个队伍的第一个特征值

team1_features =[team1_elo]

team2_features=[team2_elo]#添加我们从basketball reference.com获得的每个队伍的统计信息

for key,value inteam_stats.loc[Wteam].iteritems():

team1_features.append(value)for key,value inteam_stats.loc[Lteam].iteritems():

team2_features.append(value)#将两支队伍的特征值随机的分配在每场比赛数据的左右两侧

#并将对应的0/1赋给y值

if random.random() > 0.5:

X.append(team1_features+team2_features)

y.append(0)else:

X.append(team2_features+team1_features)

y.append(1)if skip ==0:print('X',X)

skip= 1new_winner_rank,new_loser_rank=calc_elo(Wteam,Lteam)

team_elos[Wteam]=new_winner_rank

team_elos[Lteam]=new_loser_rankreturnnp.nan_to_num(X),y#最终利用训练好的模型在 16~17 年的常规赛数据中进行预测

defpredict_winner(team_1, team_2, model):

features=[]#team 1，客场队伍

features.append(get_elo(team_1))for key, value inteam_stats.loc[team_1].iteritems():

features.append(value)#team 2，主场队伍

features.append(get_elo(team_2) + 100)for key, value inteam_stats.loc[team_2].iteritems():

features.append(value)

features=np.nan_to_num(features)returnmodel.predict_proba([features])#最终在 main 函数中调用这些数据处理函数，使用 sklearn 的Logistic Regression方法建立回归模型

if __name__=='__main__':

folder= 'data'Mstat= pd.read_csv(folder + '/15-16Miscellaneous_Stat.csv')

Ostat= pd.read_csv(folder + '/15-16Opponent_Per_Game_Stat.csv')

Tstat= pd.read_csv(folder + '/15-16Team_Per_Game_Stat.csv')

team_stats=initialize_data(Mstat, Ostat, Tstat)

result_data= pd.read_csv(folder + '/2015-2016_result.csv')

X, y=build_dataSet(result_data)#训练网络模型

print("Fitting on %d game samples.." %len(X))

model=linear_model.LogisticRegression()

model.fit(X,y)print("Doing cross-validation..")

cross_val_score(model,X,y,cv= 10,scoring='accuracy',n_jobs=-1).mean()print(model)print('Predicting on new schedule..')

schedule1617= pd.read_csv(folder + '/16-17Schedule.csv')

result=[]for index, row inschedule1617.iterrows():

team1= row['Vteam']

team2= row['Hteam']

pred=predict_winner(team1, team2, model)

prob=pred[0][0]if prob > 0.5:

winner=team1

loser=team2

result.append([winner, loser, prob])else:

winner=team2

loser=team1

result.append([winner, loser,1 -prob])

with open('16-17Result.csv', 'w') as f:

writer=csv.writer(f)

writer.writerow(['win', 'lose', 'probability'])

writer.writerows(result)print('done.')

本文链接：https://my.lmcjl.com/post/13464.html