-
Notifications
You must be signed in to change notification settings - Fork 0
/
predict_team_by_occupancy_maps.py
127 lines (110 loc) · 5.36 KB
/
predict_team_by_occupancy_maps.py
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
import numpy as np
import pandas as pd
import plotly.graph_objects as go
from sklearn import neighbors, metrics
from sklearn.decomposition import PCA
from sklearn.discriminant_analysis import LinearDiscriminantAnalysis as LDA
from sklearn.manifold import TSNE
from sklearn.preprocessing import StandardScaler
from codes.data_parsers.stats_bomb.json_directories import JsonDirectories
def predict_team_by_occupancy_maps(play_segments_length, x_bins, y_bins, train_occurances, k):
maps_to_train, maps_to_test = prepare_maps(train_occurances, play_segments_length, x_bins, y_bins)
knn = neighbors.KNeighborsClassifier(n_neighbors=k, p=2, weights = 'uniform')\
.fit(maps_to_train['occupancy_map'].tolist(), maps_to_train['team'].tolist())
pred = knn.predict(maps_to_test['occupancy_map'].tolist())
return metrics.accuracy_score(maps_to_test['team'].tolist(), pred)
def prepare_maps(train_occurances, play_segment_length, x_bins, y_bins):
json_directories = JsonDirectories()
maps = pd.read_pickle(json_directories.create_occupancy_maps_pkl_path(play_segment_length, x_bins, y_bins))
# maps = maps[filterDataByTeamsOccurancesMoreThan(list(maps['team']), 10)]
maps = maps[filterEnglishWomanTeams(list(maps['team']))]
unique_teams = getUniqueTeams(list(maps['team']))
# showDataPCA(maps, unique_teams)
# showDataTSNE(maps, unique_teams)
train_maps, test_maps = subsetTrainAndTestSet(maps, train_occurances, unique_teams)
# do_lda_transform(train_maps, test_maps, 11)
changeTeamsForIndexes(train_maps, unique_teams)
changeTeamsForIndexes(test_maps, unique_teams)
return train_maps, test_maps
def do_lda_transform(train_maps, test_maps, classes_cnt):
sc = StandardScaler()
lda = LDA(n_components=classes_cnt)
lda_train_maps = lda.fit_transform(sc.fit_transform(list(train_maps['occupancy_map'])), list(train_maps['team']))
train_maps.drop(columns=['occupancy_map'])
train_maps['occupancy_map'] = pd.DataFrame(lda_train_maps).apply(lambda r: tuple(r), axis=1).apply(np.array)
lda_test_maps = (lda.transform(sc.transform(list(test_maps['occupancy_map']))))
test_maps.drop(columns=['occupancy_map'])
test_maps['occupancy_map'] = pd.DataFrame(lda_test_maps).apply(lambda r: tuple(r), axis=1).apply(np.array)
return train_maps, test_maps
def filterDataByTeamsOccurancesMoreThan(teams, occurances_cnt):
unique_teams = getUniqueTeams(teams)
occurances = {}
for i in range(0,len(unique_teams)):
cnt = teams.count(unique_teams[i])
if cnt >= occurances_cnt and unique_teams[i] != 'FC Barcelona':
occurances[unique_teams[i]] = True
else:
occurances[unique_teams[i]] = False
return [occurances[team] for team in teams]
def filterEnglishWomanTeams(teams):
legit_teams = ['Reading WFC','West Ham United LFC','Manchester City WFC','Bristol City WFC','Arsenal WFC','Brighton & Hove Albion WFC','Everton LFC','Birmingham City WFC','Yeovil Town LFC','Chelsea FCW','Liverpool WFC']
return [team in legit_teams for team in teams]
def subsetTrainAndTestSet(data, team_train_occurances, unique_teams):
train_set = pd.DataFrame()
test_set = pd.DataFrame()
for id in unique_teams:
train_set = train_set.append(data.loc[data['team'] == id].iloc[1:team_train_occurances, ])
test_set = test_set.append(data.loc[data['team'] == id].iloc[team_train_occurances+1:, ])
return train_set, test_set
def changeTeamsForIndexes(data, unique_teams):
for index, row in data.iterrows():
row['team'] = unique_teams.index(row['team'])
def getUniqueTeams(teams):
return list(set(teams))
def showDataPCA(dataframes, unique_teams):
data = StandardScaler().fit_transform(list(dataframes['occupancy_map']))
pca = PCA(n_components=2)
principalComponents = pca.fit_transform(data)
principalDf = pd.DataFrame(data=principalComponents
, columns=['component 1', 'component 2'])
t = dataframes[['team']]
t = t.set_index(pd.Index(range(0,len(t))))
finalDf = pd.concat([t, principalDf], axis=1)
makeGraph(finalDf, unique_teams, 'PCA')
def showDataTSNE(dataframes, unique_teams):
data = list(dataframes['occupancy_map'])
tsne = TSNE(n_components=2)
principalComponents = tsne.fit_transform(data)
principalDf = pd.DataFrame(data=principalComponents
, columns=['component 1', 'component 2'])
t = dataframes[['team']]
t = t.set_index(pd.Index(range(0,len(t))))
finalDf = pd.concat([t, principalDf], axis=1)
makeGraph(finalDf, unique_teams, 'TSNE')
def makeGraph(dataframes, unique_teams, plot_title):
if(plot_title == 'PCA'):
axis_range = [-16,16]
else:
axis_range = [-30, 30]
fig = go.Figure()
fig.update_layout(
title=go.layout.Title(
text=plot_title,
xref="paper",
x=0
),
xaxis=dict(
nticks=10, range=axis_range, ),
yaxis=dict(
nticks=10, range=axis_range, )
)
targets = unique_teams
for target in targets:
indicesToKeep = dataframes['team'] == target
fig.add_trace(go.Scatter(
x=list(dataframes.loc[indicesToKeep, 'component 1']),
y=list(dataframes.loc[indicesToKeep, 'component 2']),
name=target,
mode="markers",
marker = dict(colorscale="Spectral")))
fig.show()