Projects

## required packages
import pandas as pd 
import numpy as np
import matplotlib.pyplot as plt 
import seaborn as sns 

class SoccerDataset(object):
    def __init__(self):
        self.DATA_DIR='data/SoccerData.xlsx'
        self.sheet='All Data'
        self.data=self.load_data()
    #load data
    def load_data(self):
        data = pd.read_excel(self.DATA_DIR, self.sheet, index_col=None)#reading data in particular sheet 'All Data'
        return data
    
    ##get list of column names
    def getColumns(self):
        return self.data.columns
    
    ##shows non null value caount and data type per column 
    def getDataInfo(self):
        return self.data.info()

#Loading Data
SoccerDataset=SoccerDataset()
data=SoccerDataset.data

## get columns
SoccerDataset.getColumns()

fig, ax = plt.subplots(figsize=(15, 5))
data.groupby('league').count()[['home_team']].plot.bar(
                      ax=ax,
                      title="Frequency distribution of Leagues over teams")
plt.show()

## get info about data
SoccerDataset.getDataInfo()

## to check distribution of each categorical features we should define data type to appropriate value.
##Here I am casting data type of all categorical features to int16 except RESULT column.
## Because later, I will replace the values of target column to fixed number to indicate WIN, LOSS and DRAW
data['league'] =data['league'].astype('category').cat.codes
data['home_team'] =data['home_team'].astype('category').cat.codes
data['away_team'] =data['away_team'].astype('category').cat.codes
data['date'] =data['date'].astype('category').cat.codes

## lets split data based on result feature to explore probability of winning if home or away team

data['RESULT'] = data['RESULT'].replace(['HOME'], 'WIN')#win
data['RESULT'] = data['RESULT'].replace(['AWAY'], 'LOSS')#loss
data['RESULT'] = data['RESULT'].replace(['DRAW'], 'DRAW')#draw

data['RESULT'].head()

## lets split data based on result feature to explore probability of winning if home or away team
# score_data_home = data[data['RESULT'] == 'HOME'] ##only home 
data['RESULT'] = data['RESULT'].replace(['HOME'], 'LOSS')
data['RESULT'] = data['RESULT'].replace(['AWAY'], 'WIN')
data['RESULT'] = data['RESULT'].replace(['DRAW'], 'DRAW')
data['RESULT'].head()

# data=data.loc[data['RESULT'] == 'HOME']
sns.set(rc={'figure.figsize':(10,300)})
ax=sns.countplot(y='home_team',hue='RESULT',data=data)
for p in ax.patches:
    patch_height = p.get_height()
    if np.isnan(patch_height):
        patch_height = 0
    ax.annotate('{}'.format(int(patch_height)), (p.get_x()+0.01, patch_height+3))
plt.title("Teams played at Home Vs. Chance of win/Loss/Draw")
plt.show()

sns.set(rc={'figure.figsize':(10,300)})
ax=sns.countplot(y='away_team',hue='RESULT',data=data)
for p in ax.patches:
    patch_height = p.get_height()
    if np.isnan(patch_height):
        patch_height = 0
    ax.annotate('{}'.format(int(patch_height)), (p.get_x()+0.05, patch_height+10))
plt.title("Teams played at Away Vs. Chance of win/Loss/Draw")
plt.show()

_=data.hist(figsize=(16, 20), bins=50, xlabelsize=8, ylabelsize=8)

#Using Pearson Correlation
plt.figure(figsize=(5,5))
cor=data[['home_odd','draw_odd','away_odd','RESULT']].corr()
sns.heatmap(cor, annot=True, cmap=plt.cm.Reds)
plt.show()

#Lets explore more about corelation with other final result values
#Using Pearson Correlation
plt.figure(figsize=(10,10))
cor=data[['home_odd','draw_odd','away_odd','RESULT']].corr()
sns.heatmap(cor, annot=True, cmap=plt.cm.Reds)
plt.show()

#Teams played more than 10 games
played_teams_count = data['home_team'].value_counts()
no_game_played=[]
most_played_teams=[]
for i,team in zip(played_teams_count.index,played_teams_count.values):
    if team>10:
            no_game_played.append(team)
            most_played_teams.append(i)
soccer_data=pd.DataFrame({'Team':most_played_teams,'GameCount':no_game_played})
soccer_data.tail(20) ##to check it only includes game count greater than 10. N:B: the data is decendingly sorted.

import numpy as np
corr = data.corr()
fig = plt.figure(figsize=(25,25))
ax = fig.add_subplot(111)
cax = ax.matshow(corr,cmap='coolwarm', vmin=-1, vmax=1)
fig.colorbar(cax)
ticks = np.arange(0,len(data.columns),1)
ax.set_xticks(ticks)
plt.xticks(rotation=90)
ax.set_yticks(ticks)
ax.set_xticklabels(data.columns)
ax.set_yticklabels(data.columns)
plt.show()

target_count = data.RESULT.value_counts()
print('AWAY:', target_count[0])
print('HOME:', target_count[1])
print('DRWAN:', target_count[2])
fig, ax = plt.subplots(figsize=(5, 5))
ax.tick_params(axis='x', labelsize=15)
ax.tick_params(axis='y', labelsize=10)
ax.set_xlabel('Result', fontsize=15)
ax.set_ylabel('Class Count' , fontsize=15)
ax.set_title('Distribution of traget classes (AWAY/HOME/DRAWN)', fontsize=15, fontweight='bold')
_=target_count.plot(ax=ax, kind='bar')

##As we can see from the result, data distribution for AWAY, HOME and DROWN is relatively unballanced so I have to consider 
## ballancing mechanisims appropriate to the problem. Lets use ... approach because of