华为AI Gallery社区《基于机器学习的欧洲杯赛事预测》技术分析系列——（完）预测

他山之石可以攻玉。

华为AI Gallery社区的技术贴。

传送门：

系列（一）讲述华为AI Gallery作者通过比赛基本信息为特征构建的模型框架，分析了这种框架构建的优缺点；

系列（二）中讲述了足球预测结果 二分类问题的处理以及需要掌握的知识，即所需技能Notebook+Python+Pandas。

系列（三）中讲述了建模过程分为： 1.读取数据 2.数据处理 3.数据分割 4.训练模型 5.模型评估

免费数据源1

如果您是单纯的进行数据分析，利用赔率亚盘凯利必发等找出与赛果的关系，那么可以通过免费获得即可，不要先消耗大量的时间和精力去做爬虫，等到数据齐全再去数据分析，因为数据分析也不一定会给你带来任何启示。

这期介绍赛事进行赛果预测。

3 赛果预测

本小节根据第2节对本届欧洲杯赛事进行赛果预测。

• 将第1节2015-01-01至2021-05-31的所有数据作为训练集，训练模型。
• 处理数据，将需要预测的数据按照第一节的数据处理步骤得到预测集。
• 得到预测结果。

#3.1 将2015-01-01至2021-05-31的所有数据作为训练集，训练模型 
inputs = {
    "dataframe":missing_value_impute.get_outputs()['output_port_1']   #@input {"label":"dataframe","type":"DataFrame"}
}
params = {
    "inputs": inputs,
    "column_name": "date",  #@param {"label":"column_name","type":"string","required":"true","helpTip":""}
    "condition_map_str": "BETWEEN:2015-01-01,2021-05-31"  #@param {"label":"condition_map_str","type":"string","required":"true","helpTip":""}
}
dataset_filter_all_data = MLSDatasetFilter(**params)
dataset_filter_all_data.run()
#@output {"label":"dataframe","name":"dataset_filter_all_data.get_outputs()['output_port_1']","type":"DataFrame"}


inputs = {
    "dataframe": dataset_filter_all_data.get_outputs()['output_port_1']  #@input {"label":"dataframe","type":"DataFrame"}
}
params = {
    "inputs": inputs,
    "selected_cols_str": "month,  season,  home_team,  away_team,  tournament, neutral,  win_result,num_5,diff_num_5,win_num_5,lose_num_5,num_3,diff_num_3,win_num_3,lose_num_3,num_1,diff_num_1,win_num_1,lose_num_1,\
    num_team_5,diff_num_team_5,win_num_team_5,lose_num_team_5,num_team_3,diff_num_team_3,win_num_team_3,lose_num_team_3,num_team_1,diff_num_team_1,win_num_team_1,lose_num_team_1,\
    num_year_15,diff_num_year_15,win_num_year_15,lose_num_year_15,num_year_7,diff_num_year_7,win_num_year_7,lose_num_year_7,num_year_3,diff_num_year_3,win_num_year_3,lose_num_year_3,num_year_2,diff_num_year_2,win_num_year_2,lose_num_year_2,num_year_1,diff_num_year_1,win_num_year_1,lose_num_year_1,\
    away_num,away_win_num,away_lose_num,away_win_rate,home_num,home_win_num,home_lose_num,home_win_rate"}
select_columns_all_data = MLSSelectColumns(**params)
select_columns_all_data.run()
#@output {"label":"dataframe","name":"select_columns_all_data.get_outputs()['output_port_1']","type":"DataFrame"}


inputs = {
    "dataframe": select_columns_all_data.get_outputs()['output_port_1']  #@input {"label":"dataframe","type":"DataFrame"}
}
params = {
    "inputs": inputs,
    "b_output_action": True,
    "b_use_default_encoder": True,  #@param {"label": "b_use_default_encoder", "type": "boolean", "required": "true", "helpTip": ""}
    "input_features_str": "",  #@param {"label": "input_features_str", "type": "string", "required": "false", "helpTip": ""}
    "outer_pipeline_stages": None,
    "label_col": "win_result",  #@param {"label": "label_col", "type": "string", "required": "true", "helpTip": "target label column"}
    "classifier_label_index_col": "label_index",  #@param {"label": "classifier_label_index_col", "type": "string", "required": "true", "helpTip": ""}
    "classifier_feature_vector_col": "model_features",  #@param {"label": "classifier_feature_vector_col", "type": "string", "required": "true", "helpTip": ""}
    "prediction_col": "prediction",  #@param {"label": "prediction_col", "type": "string", "required": "true", "helpTip": ""}
    "prediction_index_col": "prediction_index",  #@param {"label": "prediction_index_col", "type": "string", "required": "true", "helpTip": ""}
    "max_iter": 100,  #@param {"label": "max_iter", "type": "integer", "required": "true", "range": "(0,2147483647]", "helpTip": ""}
    "reg_param": 0,  #@param {"label": "reg_param", "type": "number", "required": "true", "range": "[0,none)", "helpTip": ""}
    "elastic_net_param": 0, #@param {"label": "elastic_net_param", "type": "number", "required": "true", "range": "[0,none)", "helpTip": ""}
    "tol": 0.000001,  #@param {"label": "tol", "type": "number", "required": "true", "range": "(0,none)", "helpTip": ""}
    "fit_intercept": True,  #@param {"label": "fit_intercept", "type": "boolean", "required": "true", "helpTip": ""}
    "standardization": True,  #@param {"label": "standardization", "type": "boolean", "required": "true", "helpTip": ""}
    "aggregation_depth": 2,  #@param {"label": "aggregation_depth", "type": "integer", "required": "true", "range": "(0,2147483647]", "helpTip": ""}
    "family": "auto",  #@param {"label": "family", "type": "enum", "required": "true", "options":"auto,binomial,multinomial", "helpTip": ""}
    "lower_bounds_on_coefficients": None,
    "upper_bounds_on_coefficients": None,
    "lower_bounds_on_intercepts": None,
    "upper_bounds_on_intercepts": None
}
lr_classifier_model = MLSLogisticRegressionClassifier(**params)
lr_classifier_model.run()
#@output {"label":"pipeline_model","name":"lr_classifier_model.get_outputs()['output_port_1']","type":"PipelineModel"}

#3.2 处理预测数据集
# 初始化预测数据test_data
# 以2021-06-17土耳其VS威尔士为例，构建测试数据
# date，home_team，away_team，tournament，neutral
test_data = {'date':'2021-06-17','home_team':'Turkey','away_team':'Wales','tournament':'UEFA Euro','neutral':'True'}
numeric_features = ['num_5','diff_num_5','win_num_5','lose_num_5','num_3','diff_num_3','win_num_3','lose_num_3','num_1','diff_num_1','win_num_1','lose_num_1','num_team_5','diff_num_team_5','win_num_team_5','num_team_3',
 'diff_num_team_3','win_num_team_3','num_team_1','diff_num_team_1','win_num_team_1','lose_num_team_5','lose_num_team_3','lose_num_team_1',
 'num_year_15','diff_num_year_15','win_num_year_15','num_year_7','diff_num_year_7','win_num_year_7','num_year_3','diff_num_year_3','win_num_year_3','num_year_2','diff_num_year_2','win_num_year_2','num_year_1','diff_num_year_1','win_num_year_1','lose_num_year_15','lose_num_year_7','lose_num_year_3','lose_num_year_2','lose_num_year_1',
 'away_num','away_win_num','away_lose_num','away_win_rate','home_num','home_win_num','home_lose_num','home_win_rate']
def init_test_data(test_data):
    for col in numeric_features:
        test_data[col]= 0.0
init_test_data(test_data)

# 生成测试数据
import numpy as np
import pandas as pd

# 读取数据
df = pd.read_csv("raw_data.csv")

# 生成测试数据
def process_test_data(test_data,df):
    df['date']= pd.to_datetime(df['date'])
    df['diff'] = df['home_score']-df['away_score']
    df['win_result'] = df['diff'].apply(lambda x: 1 if x>0 else 0)
    df = df[df['tournament']!='Friendly']
    home_team = test_data['home_team']
    away_team = test_data['away_team']
    home_team_df = df[df['home_team']==home_team]
    home_away_team_df = df[(df['home_team']==home_team) & (df['away_team']==away_team)]
    home_team_df = home_team_df.sort_values(['date'],ascending=False).reset_index() 
    home_away_team_df = home_away_team_df.sort_values(['date'],ascending=False).reset_index()
    home_team_year = int(test_data['date'].split('-')[0])
    away_team_df = df[df['away_team']==away_team]
    if(home_team_df.shape[0]>0):
        for suffix in ['5','3','1']: 
            flag = False 
            j = 0
            while j<=int(suffix)-1 and j<=home_team_df.shape[0]:     
                j+=1
                flag = True 
            j = j-1
            if flag and j>=0:
                test_data['num_'+suffix] = j+1
                test_data['diff_num_'+suffix] = home_team_df.loc[0:j,'diff'].mean()
                test_data['win_num_'+suffix]=home_team_df.loc[0:j,'win_result'].sum()  
                test_data['lose_num_'+suffix]= test_data['num_'+suffix]-test_data['win_num_'+suffix]

    if(home_away_team_df.shape[0]>0):
        for suffix in ['5','3','1']:
            j = 0
            flag = False    
            while j<=int(suffix)-1 and j<=home_away_team_df.shape[0]:       
                j+=1
                flag = True 
            j = j-1
            if flag and j<int(suffix) and j>=0:
                test_data['num_team_'+suffix]=j+1
                test_data['diff_num_team_'+suffix]=home_away_team_df.loc[0:j,'diff'].mean()
                test_data['win_num_team_'+suffix]=home_away_team_df.loc[0:j,'win_result'].sum() 
                test_data['lose_num_team_'+suffix] = test_data['num_team_'+suffix]-test_data['win_num_team_'+suffix]

    if(home_team_df.shape[0]>0):
        for suffix in ['15','7','3','2','1']:
            j = 0
            flag = False 
            while j<=home_team_df.shape[0] and home_team_year-home_team_df.loc[j,'date'].year<=int(suffix):       
                j+=1
                flag = True 
            j = j-1
            if flag: 
                test_data['num_year_'+suffix]=j+1
                test_data['diff_num_year_'+suffix] = home_team_df.loc[0:j,'diff'].mean()
                test_data['win_num_year_'+suffix] = home_team_df.loc[0:j,'win_result'].sum() 
                test_data['lose_num_year_'+suffix] = test_data['num_year_'+suffix] - test_data['win_num_year_'+suffix]              
        test_data['home_num']=home_team_df.shape[0]
        test_data['home_win_num']=home_team_df['win_result'].sum()  
        test_data['home_lose_num']= test_data['home_num']-test_data['home_win_num']
        test_data['home_win_rate']=test_data['home_win_num']/test_data['home_num']

    if away_team_df.shape[0]>0:
        test_data['away_num']=away_team_df.shape[0]
        test_data['away_lose_num'] = away_team_df['win_result'].sum()
        test_data['away_win_num'] =test_data['away_num']-test_data['away_lose_num']
        test_data['away_win_rate']=test_data['away_win_num']/test_data['away_num']
    test_data['month'] = str(int(test_data['date'].split('-')[1]))
    test_data['season'] = str(int(int(test_data['date'].split('-')[1])/3))

# 保存测试数据
process_test_data(test_data,df)
test_data.pop('date')
test_df = pd.DataFrame([test_data])
test_df.to_csv('./predict.csv')

#3.2 得到预测结果 
params = {
    "input_file_path": "./predict.csv",  #@param {"label":"input_file_path","type":"string","required":"true","helpTip":""}
    "format": "csv",  #@param {"label":"format","type":"string","required":"false","helpTip":""}
    "has_header": True,  #@param {"label":"has_header","type":"boolean","required":"false","helpTip":""}
    "delimiter": ","  #@param {"label":"delimiter","type":"string","required":"false","helpTip":""}
}
read_predict_data = MLSReadData(**params)
read_predict_data.run()
#@output {"label":"dataframe","name":"read_predict_data.get_outputs()['output_port_1']","type":"DataFrame"}


inputs = {
    "dataframe": read_predict_data.get_outputs()['output_port_1']  #@input {"label":"dataframe","type":"DataFrame"}
}
params = {
    "inputs": inputs,
    "column_type_map_str": "neutral:string,month:string,season:string"  #@param {"label":"column_type_map_str","type":"string","required":"true","helpTip":""}
}
modify_predict_data_type = MLSModifyDataType(**params)
modify_predict_data_type.run()
#@output {"label":"dataframe","name":"modify_predict_data_type.get_outputs()['output_port_1']","type":"DataFrame"}

inputs = {
    "dataframe":modify_predict_data_type.get_outputs()['output_port_1'],  #@input {"label":"dataframe","type":"DataFrame"}
    "pipeline_model": lr_classifier_model.get_outputs()['output_port_1']  #@input {"label":"pipeline_model","type":"PipelineModel"}
}
params = {
    "inputs": inputs
}
model_predict = MLSModelPredict(**params)
model_predict.run()
#@output {"label":"dataframe","name":"model_predict.get_outputs()['output_port_1']","type":"DataFrame"}

inputs = {
    "dataframe": model_predict.get_outputs()['output_port_1']  #@input {"type":"DataFrame", "label": "dataframe"}
}
p_df = inputs["dataframe"].toPandas()
p_df[['home_team','away_team','prediction']]

     home_team     away_team     prediction
     Turkey              Wales                   0
#该预测结果为Wales胜利。