3244/player_model + time.py at main · shuyang0/3244 · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
#analyse diff algos --> kNN should not perform well
#https://scikit-learn.org/stable/modules/svm.html
#https://intellipaat.com/blog/tutorial/machine-learning-tutorial/svm-algorithm-in-python/#How-Does-Support-Vector-Machine-Work


### X labels: 2 x 11 player ratings ###
import datetime
import pandas as pd
from sklearn.model_selection import train_test_split
from sklearn.metrics import *
from sklearn.model_selection import cross_validate, KFold
from helper import *

algo = input(algos_available)
print('\n')
label = input(labels_available)
print('\n')
scorers = input(scorers_available)
print('\nPlease wait...\n')

#suppress warnings
import warnings
warnings.filterwarnings("ignore")

#read data from Match.csv, extract columns into pandas dataframe
data = pd.read_csv (r"Match.csv")
matches = pd.DataFrame(data , columns= ["league_id", "date", "home_team_goal","away_team_goal", \
                                        "home_player_1","home_player_2","home_player_3",\
                                        "home_player_4","home_player_5","home_player_6",\
                                        "home_player_7","home_player_8","home_player_9",\
                                        "home_player_10","home_player_11","away_player_1",\
                                        "away_player_2","away_player_3","away_player_4",\
                                        "away_player_5","away_player_6","away_player_7",\
                                        "away_player_8","away_player_9","away_player_10",\
                                        "away_player_11"])

#drop rows with missing information
matches = matches.dropna()

#convert all values to int
datatype_dict = {}
for column in matches:
    if column == "date":
        datatype_dict["date"] = "datetime64"
    else:
        datatype_dict[column] = "Int64"
matches = matches.astype(datatype_dict)

#take only EPL matches (league_id  == 1729)
matches = matches.loc[matches["league_id"] == 1729 ]

#read data from Player_Attributes.csv, extract columns into pandas dataframe
data = pd.read_csv (r"Player_Attributes.csv")
players = pd.DataFrame(data , columns= ["player_api_id", "date", "overall_rating"])

#drop rows with missing information
players = players.dropna()

#convert to appropriate data types
players = players.astype({
    "date": "datetime64",
    "player_api_id": "Int64",
    "overall_rating": "Int64",
    })

#create players_dict, which maps {player_api_id -> {date -> rating} }
players_dict = {}
for index, row in players.iterrows():
    if row["player_api_id"] not in players_dict:
        players_dict[row["player_api_id"]] = {row["date"]:row["overall_rating"]}
    else:
        players_dict[row["player_api_id"]][row["date"]] = row["overall_rating"]

# Mapping function that takes in a row + dictionary containing match date and player_id, and returns
# the player overall rating that is the closest one before the input match date

def apply_rating(row, col_name):
    match_date = row["date"]
    player_id = row[col_name]
    ratings = list(players_dict[player_id].items())
    ratings = filter(lambda x: x[0] <= match_date, ratings)
    return sorted(ratings, key = lambda x: x[0], reverse = True)[0][1]


# dataframe.apply with axis=1 iterates through the rows of the dataframe, applying the higher order function to the row
# ("x" in this case).
# The returned player overall score is stored under the new column "home_player_x_score" or "away_player_x_score" respectively.
for i in range(1,12):
    matches[f"home_player_{i}_rating"] = matches.apply(
        lambda x: apply_rating(x, f"home_player_{i}"),
        axis=1
    )
    matches[f"away_player_{i}_rating"] = matches.apply(
        lambda x: apply_rating(x, f"away_player_{i}"),
        axis=1
    )

matches, label_name = getLabel(label, matches)

scorers = getScorers(scorers)

#remove columns not used in ML model
matches = matches.drop("league_id", axis = 1)
matches = matches.drop("home_team_goal", axis = 1)
matches = matches.drop("away_team_goal", axis = 1)
matches = matches.drop("date", axis = 1)
for i in range(1,12):
    matches.drop(f"home_player_{i}", axis = 1)
    matches.drop(f"away_player_{i}", axis = 1)

#print statements
#print(matches)
#with pd.option_context("display.max_rows", 20, "display.max_columns", None):  print(matches)


#MODEL TRAINING

#None: single core CV, -1: all cores CV
multicore = None

#X-y split
X = matches.drop('result', axis = 1)
y = matches['result']

#train model with 10-fold cross validation, using on given algo
kfold = KFold(n_splits=10, shuffle=True, random_state=420)
#algos = ('svm', 'knn', 'nb', 'dt', 'rf', 'lr', 'p')
algos = ('svm', 'knn', 'nb', 'dt', 'rf', 'lr')
if algo == 'all':
    for a in algos:
        model, algo_name = getAlgo(a)
        scores = cross_validate(model, X, y, cv=kfold, scoring = scorers, n_jobs = multicore)
        print(algo_name + ' to predict ' + label_name + ' using player ratings(time adjusted)')
        if 'f1_weighted' in scorers:
            print('F1-weighted: ' + str(round(scores['test_f1_weighted'].mean(), 4)) + ' (+/-' + str(round(scores['test_f1_weighted'].std(), 4)) + ')')
        if 'accuracy' in scorers:
            print('Accuracy: ' + str(round(scores['test_accuracy'].mean(), 4)) + ' (+/-' + str(round(scores['test_accuracy'].std(), 4)) + ')')
        if 'roc_auc_ovo_weighted' in scorers:
            print('AUC-ROC: ' + str(round(scores['test_roc_auc_ovo_weighted'].mean(), 4)) + ' (+/-' + str(round(scores['test_roc_auc_ovo_weighted'].std(), 4)) + ')')
        print('\n')
else:
    model, algo_name = getAlgo(algo)
    scores = cross_validate(model, X, y, cv=kfold, scoring = scorers, n_jobs = multicore)
    print(algo_name + ' to predict ' + label_name + ' using player ratings (time adjusted)')
    if 'f1_weighted' in scorers:
        print('F1-weighted: ' + str(round(scores['test_f1_weighted'].mean(), 4)) + ' (+/-' + str(round(scores['test_f1_weighted'].std(), 4)) + ')')
    if 'accuracy' in scorers:
        print('Accuracy: ' + str(round(scores['test_accuracy'].mean(), 4)) + ' (+/-' + str(round(scores['test_accuracy'].std(), 4)) + ')')
    if 'roc_auc_ovo_weighted' in scorers:
        print('AUC-ROC: ' + str(round(scores['test_roc_auc_ovo_weighted'].mean(), 4)) + ' (+/-' + str(round(scores['test_roc_auc_ovo_weighted'].std(), 4)) + ')')

#train model with 80-20 train-test split, using on given algo
#X_train, X_test, y_train,  y_test = train_test_split(X, y, test_size = 0.20)
#model, algo_name = getAlgo(a)
#model.fit(X_train, y_train)
#y_pred = model.predict(X_test)
#print(algo_name + ' to predict ' + label_name + ' using player ratings' + '\n')
#print(confusion_matrix(y_test, y_pred))
#print(classification_report(y_test, y_pred))