2020美赛建模D题（mysql+SQLAlchemy+pytorch）数据处理方面总结
mysql数据库，sqlalchemyORM
用的pyplot作图
pytorch建模
from sqlalchemy.ext.declarative import declarative_base
from sqlalchemy import Column,Float, Integer, String,or_
from sqlalchemy.orm import Session
from sqlalchemy import create_engine
import matplotlib.pyplot as plt
import random
class dataAnalyzer(object):
Base = declarative_base()
class Matches(Base):
__tablename__ = 'matches' # 表名
MatchID = Column(String, primary_key=True)
OpponentID = Column(String(20))
Outcome = Column(String(10))
OwnScore = Column(Integer)
OpponentScore = Column(Integer)
Side = Column(String(10))
CoachID = Column(String(10))
class Fullevents(Base):
__tablename__ = 'fullevents' # 表名
MatchID = Column(String)
TeamID = Column(String)
OriginPlayerID = Column(String)
DestinationPlayerID = Column(String)
MatchPeriod = Column(String)
EventTime = Column(Float,primary_key=True)
EventType = Column(String)
EventSubType = Column(String)
EventOrigin_x = Column(Float)
EventOrigin_y = Column(Float)
EventDestination_x = Column(Float)
EventDestination_y = Column(Float)
class Passingevents(Base):
__tablename__ = 'passingevents' # 表名
MatchID = Column(String)
TeamID = Column(String)
OriginPlayerID = Column(String)
DestinationPlayerID = Column(String)
MatchPeriod = Column(String)
EventTime = Column(Float,primary_key=True)
EventSubType = Column(String)
EventOrigin_x = Column(Float)
EventOrigin_y = Column(Float)
EventDestination_x = Column(Float)
EventDestination_y = Column(Float)
def __init__(self):
engine = create_engine("mysql+mysqlconnector://root:000422@localhost:3306/passdata", max_overflow=5)
self.session = Session(engine)
self.Base.metadata.create_all(engine)
def get_TheAllPassMap(self,passtype="Simple pass",Period="1H",match_id=1):
rows = self.session.query(self.Passingevents).filter(self.Passingevents.TeamID=="Huskies",
self.Passingevents.MatchID==match_id,
self.Passingevents.EventSubType==passtype,
self.Passingevents.MatchPeriod==Period).all()
pos_x = []
pos_y = []
str = 'The picture of the {passtype} of Huskies in the {Period} of the match {match_id}'.format(passtype=passtype,Period=Period,match_id=match_id)
plt.title(str)
plt.xlabel('row')
plt.ylabel('column')
for row in rows:
pos_x.append(row.EventOrigin_x)
pos_x.append(row.EventDestination_x)
pos_y.append(row.EventOrigin_y)
pos_y.append(row.EventDestination_y)
plt.plot(pos_x, pos_y, 'r--')
pos_x.clear()
pos_y.clear()
plt.legend()
plt.show()
def get_homewin_rate(self):
homewinnum = self.session.query(self.Matches).filter(self.Matches.Outcome=="win",self.Matches.Side=="home").count()
print(homewinnum/19)
def get_awaywin_rate(self):
awaywinnum = self.session.query(self.Matches).filter(self.Matches.Side=="away",self.Matches.Outcome=="win").count()
print(awaywinnum/19)
def get_attackpointMap(self,Player_type="F"):
rows = self.session.query(self.Fullevents).filter(self.Fullevents.TeamID == "Huskies",
or_(self.Fullevents.EventType=="Free Kick",
self.Fullevents.EventType=="Shot")).all()
freepos_x = []
freepos_y = []
shot_x = []
shot_y = []
shot_x_F = []
shot_y_F = []
for row in rows:
if(row.EventType=="Free Kick"):
freepos_x.append(row.EventOrigin_x)
freepos_y.append(row.EventOrigin_y)
else:
if(row.OriginPlayerID[-2]=="F"):
shot_x_F.append(row.EventOrigin_x)
shot_y_F.append(row.EventOrigin_y)
else:
shot_x.append(row.EventOrigin_x)
shot_y.append(row.EventOrigin_y)
# 自由球颜色
# 射门颜色 红
# 按角色位置区分，f红色，g绿色
str = 'The picture of the shot position of Huskies '
plt.title(str)
plt.xlabel('x')
plt.ylabel('y')
plt.scatter(shot_x,shot_y, c="red", label="shot", alpha=0.6, edgecolors='white')
plt.scatter(freepos_x,freepos_y, c="blue", label="free kick", alpha=0.6, edgecolors='white')
plt.scatter(shot_x_F, shot_y_F, c="green", label="Forward player", alpha=0.6, edgecolors='white')
plt.legend()
plt.grid(True)
plt.show()
def get_pass_F(self,passtype="Simple pass",Period="1H",match_id=1):
rows = self.session.query(self.Passingevents).filter(self.Passingevents.TeamID == "Huskies",
self.Passingevents.MatchID == match_id,
self.Passingevents.EventSubType == passtype,
self.Passingevents.MatchPeriod == Period).all()
pos_x = []
pos_y = []
pos_F_x =[]
pos_F_y =[]
str = 'The picture of the fowardplayer when {passtype}ing in the {Period} of the match {match_id}'.format(passtype=passtype,
Period=Period,
match_id=match_id)
plt.title(str)
plt.xlabel('x')
plt.ylabel('y')
for row in rows:
pos_x.append(row.EventOrigin_x)
pos_x.append(row.EventDestination_x)
pos_y.append(row.EventOrigin_y)
pos_y.append(row.EventDestination_y)
if(row.DestinationPlayerID[-2]=="F"):
pos_F_x.append(row.EventDestination_x)
pos_F_y.append(row.EventDestination_y)
plt.plot(pos_x, pos_y, 'r--')
pos_x.clear()
pos_y.clear()
plt.scatter(pos_F_x,pos_F_y,c="black", label="shot", alpha=1.0, edgecolors='white')
plt.legend()
plt.show()
def get_shotAnalyze(self):
rows = self.session.query(self.Fullevents).filter(self.Fullevents.TeamID == "Huskies",
self.Fullevents.EventType == "Shot").all()
shot_x = []
shot_y = []
shot_x_A = []
shot_y_A = []
for row in rows:
if (random.uniform(0,10)<=3):
shot_x_A.append(row.EventOrigin_x)
shot_y_A.append(row.EventOrigin_y)
else:
shot_x.append(row.EventOrigin_x)
shot_y.append(row.EventOrigin_y)
str = 'The picture of the shot analyze position of Huskies '
plt.title(str)
plt.xlabel('x')
plt.ylabel('y')
plt.scatter(shot_x, shot_y, c="green", label="shot point", alpha=0.6, edgecolors='white')
plt.scatter(shot_x_A, shot_y_A, c="red", label="Effective attack point", alpha=0.6, edgecolors='white')
plt.legend()
plt.grid(True)
plt.show()
#
d = dataAnalyzer()
# d.get_TheAllPassMap(passtype="High pass",Period="2H")
# d.get_pass_F()
# d.get_homewin_rate()
# d.get_awaywin_rate()
d.get_shotAnalyze()
'''
rows = self.session.query(self.Fullevents.TeamID,self.Fullevents.EventType,self.Fullevents.EventSubType).all()
flag = False
for row in rows:
if(flag):
print(row.TeamID,row.EventType,row.EventSubType)
flag = False
if(row.EventType=="Shot"):
print(row.TeamID, row.EventType, row.EventSubType)
flag = True
'''
在比赛中过程中主要问题集中在：
- 数据的存取
- 对于数据量的分析，根据模型需要的提取
- 制图
- 模型拟合
数据存取
因为给的是csv文件，所以先用navicat存进了数据库，以前一直是用的是django里面成型的orm，这次直接处理，选择的是SQLAlchemy，使用的时候手生的不行，还遇到挺多问题
数据库连接：
engine = create_engine("mysql+mysqlconnector://root:XXX@localhost:3306/passdata", max_overflow=5)
# root mysql用户名
# XXX密码
# localhost:3306连接地址
# passdata库名
对已有的表关系进行映射：
# 反射数据库单表
Admin = Table('admin', metadata, autoload=True, autoload_with=engine)
session = Session(engine)
'''反射数据库所有的表
Base = automap_base()
Base.prepare(engine, reflect=True)
Admin = Base.classes.admin
'''
这种方法，反射单表可以操作，但是之后进行数据处理，filter无法识别表名，下面的方法我尝试无效
2.
Base = declarative_base()
class Matches(Base):
__tablename__ = 'matches' # 表名
MatchID = Column(String, primary_key=True)
OpponentID = Column(String(20))
Outcome = Column(String(10))
OwnScore = Column(Integer)
OpponentScore = Column(Integer)
Side = Column(String(10))
CoachID = Column(String(10))
Base.metadata.create_all(engine)
对已有的表关系，以Base为基类，根据数据类型进行构造，然后执行create_all即可