-
Notifications
You must be signed in to change notification settings - Fork 0
/
summary_stats.py
299 lines (253 loc) · 14.4 KB
/
summary_stats.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
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
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
# Analyzing the Crystal Island Affect data for UMUAI Special Issue on Affect due 3/15/2018
import numpy as np
import pandas as pd
import tensorflow as tf
from sklearn.preprocessing import scale
from sklearn.linear_model import LinearRegression
from sklearn.linear_model import Ridge
from sklearn.linear_model import MultiTaskLasso
from sklearn.model_selection import cross_val_score
from sklearn.model_selection import KFold
from sklearn.linear_model import Lasso
from sklearn.neural_network import MLPRegressor
from sklearn.ensemble import RandomForestRegressor
from sklearn.neighbors import KNeighborsRegressor
from sklearn.neural_network import MLPClassifier
from sklearn.svm import SVC
from sklearn.tree import DecisionTreeClassifier
from sklearn.ensemble import RandomForestClassifier
from sklearn.naive_bayes import GaussianNB
from sklearn.neighbors import KNeighborsClassifier
from sklearn.linear_model import LogisticRegression
from sklearn.linear_model import LogisticRegressionCV
from sklearn.metrics import confusion_matrix
def calculate_vif(data, cols, col):
lm = LinearRegression()
t_list = list(cols)
t_list.remove(col)
lm.fit(data[t_list],data[col])
r2 = lm.score(data[t_list],data[col])
vif = 1. / (1 - r2)
return vif
def overall_summary(data, show=True):
if show:
print("Number of students: %d" % data.shape[0])
def response_summaries(data, response_list, show=True):
for response in response_list:
if show:
print(data[response].describe())
print("Positive(%%): %d (%.3f%%)" % (np.sum(data[response] > 0), np.mean(data[response] > 0)*100))
def independent_summaries(data, cols, show=True):
if show:
print(data[cols].describe())
print(data[cols].corr(method='pearson'))
for col in cols:
col_vif = calculate_vif(data, cols, col)
print("%s VIF: %.4f" % (col, col_vif))
def composite_tests(data, response_name, ind_names, show=True):
response_col = "%sEvidence-0-Sum" % response_name
ind_cols = ["%sEvidence-0-Sum" % e for e in ind_names]
lm = LinearRegression()
lm.fit(data[ind_cols], data[response_col])
r2 = lm.score(data[ind_cols], data[response_col])
if show:
print("%s from AUs %s R2: %.3f" % (response_name, ind_names, r2))
for c, n in zip(lm.coef_, ind_names):
print("%s: %.3f" % (n, c))
def all_summary_statistics(data, show=True):
AU_cols = [e for e in data.columns.values if "AU" in e]
Comp_cols = [e for e in data.columns.values if "AU" not in e and "Evidence" in e]
Game_cols = [e for e in data.columns.values if "Duration" in e or "C-" in e or "CD-" in e]
print(
"----------------------------------------------\n|Action Unit Summaries\n-------------------------------------------")
# Outputting analysis for paper writing
overall_summary(data, show)
response_summaries(data, response_list=["NormalizedLearningGain", "Presence"], show=show)
independent_summaries(data, cols=AU_cols, show=show)
print(
"----------------------------------------------\n|Composite Summaries\n-------------------------------------------")
independent_summaries(data, cols=Comp_cols, show=show)
print(
"----------------------------------------------\n|Game Feature Summaries\n-------------------------------------------")
independent_summaries(data, cols=Game_cols, show=show)
print(
"----------------------------------------------\n|Composites from AUs\n-------------------------------------------")
# Outputting breakdown of composites by contributing AU linear model
composite_tests(data, response_name="Joy", ind_names=["AU6", "AU12"], show=show)
composite_tests(data, response_name="Sadness", ind_names=["AU1", "AU4", "AU15"], show=show)
composite_tests(data, response_name="Surprise", ind_names=["AU1", "AU2", "AU5", "AU26"], show=show)
composite_tests(data, response_name="Fear", ind_names=["AU2", "AU4", "AU5", "AU7", "AU20", "AU26"], show=show)
composite_tests(data, response_name="Anger", ind_names=["AU4", "AU5", "AU7", "AU23"], show=show)
composite_tests(data, response_name="Disgust", ind_names=["AU9", "AU15"], show=show)
composite_tests(data, response_name="Contempt", ind_names=["AU12", "AU14"], show=show)
def add_interaction_terms(data, col_set1, col_set2):
new_data = data.copy()
col_list = []
for col in col_set1:
for col2 in col_set2:
new_data["%s-%s" % (col[:5], col2[:5])] = data[col] * data[col2]
new_data["%s-%s" % (col[:5], col2[:5])] = (new_data["%s-%s" % (col[:5], col2[:5])] - new_data["%s-%s" % (col[:5],col2[:5])].mean())/(new_data["%s-%s" % (col[:5],col2[:5])].std())
col_list.append("%s-%s" % (col[:5], col2[:5]))
return new_data, col_list
def cv_regression(model, data, feature_cols, response_col, show=False):
kf = KFold(n_splits=data.shape[0])
rss = 0
for train, test, in kf.split(data):
model.fit(X=data.loc[train, feature_cols], y=data.loc[train, response_col])
pred = model.predict(X=data.loc[test, feature_cols])[0]
rss += (pred - data.loc[test, response_col].iloc[0]) ** 2
tss = np.sum((data[response_col] - data[response_col].mean())**2)
cvR2 = 1 - rss / tss
model.fit(data.loc[:,feature_cols], data.loc[:, response_col])
r2 = model.score(X=data.loc[:, feature_cols], y=data.loc[:, response_col])
if show:
print("CVR2: %.4f" % np.mean(cvR2))
print("R2: %.4f" % r2)
# for c, n in zip(model.coef_, feature_cols):
# if abs(c) > 0.0001:
# print("%s: %.5f" % (n, c))
return cvR2, r2
def cv_classification(model, data, feature_cols, response_col, show=False):
kf = KFold(n_splits=data.shape[0])
total_correct = 0
y_pred = []
y_true = []
for train, test in kf.split(data):
model.fit(X=data.loc[train, feature_cols], y=data.loc[train, response_col])
pred = model.predict(X=data.loc[test, feature_cols])[0]
total_correct += int(np.equal(pred, data.loc[test, response_col].iloc[0]))
y_pred.append(pred)
y_true.append(data.loc[test, response_col].iloc[0])
if show:
print(confusion_matrix(y_pred=y_pred, y_true=y_true))
test_accuracy = float(total_correct) / data.shape[0]
train_accuracy = model.score(X=data.loc[:,feature_cols], y=data.loc[:,response_col])
return test_accuracy, train_accuracy
def ens_classification(model, data, feature_sets, response_col, show=False):
kf = KFold(n_splits=data.shape[0])
total_correct = 0
y_pred = []
y_true = []
for train, test in kf.split(data):
pred_probs = np.array([0] * len(data[response_col].unique()),dtype='float64')
for f_set in feature_sets:
model.fit(X=data.loc[train, f_set], y=data.loc[train, response_col])
pred_probs += model.predict_proba(X=data.loc[test, f_set])[0]
pred = np.argmax(pred_probs)
total_correct += int(np.equal(pred, data.loc[test, response_col].iloc[0]))
y_pred.append(pred)
y_true.append(data.loc[test, response_col].iloc[0])
test_accuracy = float(total_correct) / data.shape[0]
return test_accuracy, -1.0
def multivariate_regression(output_filename):
regression_output = open(output_filename, 'w')
lm = MultiTaskLasso(alpha=0.1)
reg_name = "MTLassoRegression"
gcvr2, gr2 = cv_regression(lm, n_data, Game_cols, ["NormalizedLearningGain", "Presence"], show=True)
gccvr2, gcr2 = cv_regression(lm, n_data, Game_cols + Comp_cols, ["NormalizedLearningGain", "Presence"], show=True)
gaucvr2, gaur2 = cv_regression(lm, n_data, Game_cols + AU_cols, ["NormalizedLearningGain", "Presence"], show=True)
def generate_binary_cols(df, cols):
new_cols = []
for col in cols:
df["%s-Binary" % col] = np.array(df[col] > df[col].median(), dtype=int)
new_cols.append("%s-Binary" % col)
return df, new_cols
def classification(output_filename):
classification_output = open(output_filename, 'w')
classification_output.write("Classifier,ClassBreakdown,GameTest,GameTrain,CompTest,CompTrain,AUTest,AUTrain,BCompTest,BCompTrain,BAUTest,BAUTrain\n")
classifiers = [LogisticRegression(C=0.1),
LogisticRegression(C=3.0),
SVC(C=0.1, kernel='linear'),
GaussianNB()
]
classifier_names = ["LogReg.1", "LogReg3", "SVC", "NB"]
for cl, c_name in zip(classifiers, classifier_names):
for class_breakdown in ["MedianNLG", "MedianPres", "FourClass", "ThreeClass"]:
feature_accuracies = []
for f_set in [Game_cols, Game_cols + Comp_cols, Game_cols + AU_cols, Game_cols + Comp_cols + AU_cols]:
test_acc, train_acc = cv_classification(cl, n_data, f_set, class_breakdown, show=show)
feature_accuracies.append("%.3f" % test_acc)
feature_accuracies.append("%.3f" % train_acc)
classification_output.write("%s,%s,%s\n" % (c_name, class_breakdown,",".join(feature_accuracies)))
# for cl, c_name in zip(classifiers, classifier_names):
# for class_breakdown in ["MedianNLG", "MedianPres", "FourClass", "ThreeClass"]:
# feature_accuracies = []
# f_set = [Game_cols + Comp_cols, Game_cols + AU_cols, Game_cols + bin_Comp, Game_cols + bin_AU]
# test_acc, train_acc = ens_classification(cl, n_data, f_set, class_breakdown, show=show)
# feature_accuracies.append("%.3f" % test_acc)
# feature_accuracies.append("%.3f" % train_acc)
# classification_output.write("%s,%s,%s\n" % (c_name, class_breakdown,",".join(feature_accuracies)))
def univariate_regression(output_filename):
regression_output = open(output_filename, 'w')
regression_output.write(
"ModelName,Response,GameplayCVR2,CompositeCVR2,Composite-InteractionCVR2,AUCVR2,AU-InteractionCVR2\n")
regressors = [LinearRegression(), Lasso(alpha=0.1), Ridge(alpha=1.0),
MLPRegressor(hidden_layer_sizes=(20, 5), alpha=0.0001, early_stopping=True),
RandomForestRegressor(), KNeighborsRegressor(n_neighbors=7, weights='distance')]
regressor_names = ["LinearRegression", "Lasso", "Ridge", "MLP", "RandomForestReg", "KNN"]
for lm, r_name in zip(regressors, regressor_names):
show = r_name != "MLP" and r_name != "RandomForestReg" and r_name != "KNN"
print(
"------------------------------------\nPredicting NLG from Game Cols using %s\n----------------------------------------" % r_name)
gr2 = cv_regression(lm, n_data, Game_cols, "NormalizedLearningGain", show=show)
print(
"------------------------------------\nPredicting NLG from Game Cols + Comp Cols using %s\n------------------------------" % r_name)
gc_data, gc_cols = add_interaction_terms(n_data, Game_cols, Comp_cols)
cir2 = cv_regression(lm, gc_data, gc_cols, "NormalizedLearningGain", show=show)
cr2 = cv_regression(lm, n_data, Game_cols + Comp_cols, "NormalizedLearningGain", show=show)
print(
"------------------------------------\nPredicting NLG from Game Cols + AU Cols using %s\n----------------------------------------" % r_name)
gau_data, gau_cols = add_interaction_terms(n_data, Game_cols, AU_cols)
auir2 = cv_regression(lm, gau_data, gau_cols, "NormalizedLearningGain", show=show)
aur2 = cv_regression(lm, n_data, Game_cols + AU_cols, "NormalizedLearningGain", show=show)
line = "%s,NLG,%.3f,%.3f,%.3f,%.3f,%.3f" % (r_name, gr2, cr2, cir2, aur2, auir2)
regression_output.write(line + "\n")
print(
"------------------------------------\nPredicting Presence from Game Cols using %s\n----------------------------------------" % r_name)
gr2 = cv_regression(lm, n_data, Game_cols, "Presence", show=show)
print(
"------------------------------------\nPredicting Presence from Game Cols + Comp Cols using %s\n------------------------------" % r_name)
cir2 = cv_regression(lm, gc_data, gc_cols, "Presence", show=show)
cr2 = cv_regression(lm, n_data, Game_cols + Comp_cols, "Presence", show=show)
print(
"------------------------------------\nPredicting Presence from Game Cols + AU Cols using %s\n----------------------------------------" % r_name)
auir2 = cv_regression(lm, gau_data, gau_cols, "Presence", show=show)
aur2 = cv_regression(lm, n_data, Game_cols + AU_cols, "Presence", show=show)
line = "%s,Presence,%.3f,%.3f,%.3f,%.3f,%.3f" % (r_name, gr2, cr2, cir2, aur2, auir2)
regression_output.write(line + "\n")
if __name__ == "__main__":
# Script parameter settings
show = True
# Reading and cleaning data
data_filename = "Data/Positive_Spike_SummaryPost_Appended_Std_Full.csv"
data = pd.read_csv(data_filename)
include_partial = False
if not include_partial:
cond = data["Condition"]
keep_rows = data["Condition"] == "Full"
data = data.loc[keep_rows,:]
data.drop(["TestSubject", "Condition"], axis=1, inplace=True)
n_data = data.apply(lambda col: (col - np.mean(col)) / np.std(col))
AU_cols = [e for e in data.columns.values if "AU" in e]
Comp_cols = [e for e in data.columns.values if "AU" not in e and "Evidence" in e]
Game_cols = [e for e in data.columns.values if "Duration" in e or "C-" in e or "CD-" in e]
n_data, bin_Comp = generate_binary_cols(n_data, Comp_cols)
n_data, bin_AU = generate_binary_cols(n_data, AU_cols)
if include_partial:
n_data["nCondition"] = np.array(cond == "Full", dtype=int)
Game_cols.append("nCondition")
# Summary statistics
#all_summary_statistics(data, show=show)
# Regression Models
#univariate_regression(output_filename="RegressionResults.csv")
#multivariate_regression(output_filename="MVRegressionResults.csv")
# Classification Models
n_data["MedianNLG"] = np.array(n_data["NormalizedLearningGain"] > n_data["NormalizedLearningGain"].median(), dtype=int)
n_data["MedianPres"] = np.array(n_data["Presence"] > n_data["Presence"].median(), dtype=int)
n_data["FourClass"] = n_data["MedianNLG"]*2 + n_data["MedianPres"]
n_data["ThreeClass"] = n_data["MedianNLG"] + n_data["MedianPres"]
print(n_data["FourClass"].value_counts())
print(max(n_data["FourClass"].value_counts())/n_data.shape[0])
print(n_data["ThreeClass"].value_counts())
print(max(n_data["ThreeClass"].value_counts())/n_data.shape[0])
classification(output_filename="ClassificationResultsPost.csv")