forked from potatochip/meatball_stats
-
Notifications
You must be signed in to change notification settings - Fork 0
/
parmesan.py
299 lines (249 loc) · 11.2 KB
/
parmesan.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
'''
meatball helper module
'''
import pandas as pd
from prettytable import PrettyTable
import datetime
from collections import defaultdict, Counter
import pickle
from prettytable import PrettyTable
import itertools
import matplotlib.pyplot as plt
import seaborn as sns
import numpy as np
from sklearn.feature_selection import RFE
from sklearn.ensemble import ExtraTreesClassifier
from sklearn.cross_validation import train_test_split
from sklearn.linear_model import LinearRegression
from sklearn.neighbors import KNeighborsClassifier
from sklearn.linear_model import LogisticRegression
from sklearn.naive_bayes import GaussianNB
from sklearn.svm import SVC
from sklearn.tree import DecisionTreeClassifier
from sklearn.ensemble import RandomForestClassifier
# print dataframe to screen with enough room
# pd.set_option('display.height', 1000)
pd.set_option('display.max_rows', 1000)
pd.set_option('display.max_columns', 1000)
pd.set_option('display.width', 1000)
def get_sample_dataset(dataset='processed.cleveland.data'):
'''
uses sample dataset from cleveland heart disease study if no dataset passed
'''
df = pd.DataFrame.from_csv(dataset, header=-1, index_col=None)
df.columns = ['age', 'sex', 'chest_pain', 'resting_bp', 'cholesterol',
'blood_sugar', 'ecg', 'max_hr', 'exercise_induced_angina',
'st_depression', 'slope', 'num_major_vessels', 'thal',
'diagnosis']
df.index.names = ['patient']
df = df.convert_objects(convert_numeric=True)
# changing diagnosis from 0-4 scale to just 0 or 1
df.diagnosis = df.diagnosis.apply(lambda x: 0 if x == 0 else 1)
df.dropna(inplace=True)
features = df.drop('diagnosis', axis=1)
response = df.diagnosis
return features, response, df
def data_processor(column_labels, response_label, data_file, header, index):
if header and index:
df = pd.DataFrame.from_csv(data_file, header=0, index_col=0)
else:
df = pd.DataFrame.from_csv(data_file, header=-1, index_col=None)
df.columns = column_labels
# # additional processing
# df.family_history = pd.get_dummies(df.family_history).Present
df = df.convert_objects(convert_numeric=True)
df.dropna(inplace=True)
features = df.drop(response_label, axis=1)
response = df[response_label]
return features, response, df
def recover_pickle(pickle, filename):
dt = str(datetime.datetime.now())
filename = filename + '-' + dt
df = pd.read_pickle(pickle)
df2 = df.copy()
# save with multi-index .csv
df.set_index(['Feature', 'Estimator'], inplace=True)
df.to_csv(filename+'.csv')
# save without multi-index as .txt and print to screen
pt = PrettyTable()
for i in df2.columns:
pt.add_column(i, df2[i].tolist())
print pt
table_txt = pt.get_string()
with open(filename+'.txt', 'w') as file:
file.write(table_txt)
def deep_sea_squid(estimator, parameters, X_train, X_test, y_train, y_test,
model_name, tuning='accuracy'):
# call squid grid again with narrower and more detailed parameters once model selected
pass
def max_scores(dataframe, tuning=False):
max_evaluators = defaultdict(int)
evaluators = ['Accuracy', 'Precision', 'Recall', 'F1', 'AUC']
if not tuning:
pass
else:
evaluators = [tuning]
for evaluator in evaluators:
# idx = dataframe.groupby(['Feature'])[evaluator].transform(max) == dataframe[evaluator]
# max_evaluators[evaluator] = dataframe[idx][['Feature', 'Estimator', evaluator, evaluator+'_best']]
max_evaluators = dataframe.sort(evaluator, ascending=False).groupby('Feature').first()[['Estimator', evaluator, evaluator+'_best']]
return max_evaluators
def load_pickle(pickle):
df = pd.read_pickle(pickle)
return df
def make_plot(X_train, y_train, X, y, test_data, model, model_name, features, response):
feature = X.columns
f, ((ax1, ax2), (ax3, ax4)) = plt.subplots(2, 2, sharey=False)
sns.regplot(X[feature[4]], y, test_data, ax=ax1)
sns.boxplot(X[feature[4]], y, color="Blues_r", ax=ax2)
model.fit(X_train, y_train)
sns.residplot(X[feature[4]], (model.predict(X) - y) ** 2, color="indianred", lowess=True, ax=ax3)
if model_name is 'linear':
sns.interactplot(X[feature[3]], X[feature[4]], y, ax=ax4, filled=True, scatter_kws={"color": "dimgray"}, contour_kws={"alpha": .5})
elif model_name is 'logistic':
pal = sns.blend_palette(["#4169E1", "#DFAAEF", "#E16941"], as_cmap=True)
levels = np.linspace(0, 1, 11)
sns.interactplot(X[feature[3]], X[feature[4]], y, levels=levels, cmap=pal, logistic=True)
else:
pass
ax1.set_title('Regression')
ax2.set_title(feature[4]+' Value')
ax3.set_title(feature[4]+' Residuals')
ax4.set_title('Two-value Interaction')
f.tight_layout()
plt.savefig(model_name+'_'+feature[4], bbox_inches='tight')
# Multi-variable correlation significance level
f, ax = plt.subplots(figsize=(10, 10))
cmap = sns.blend_palette(["#00008B", "#6A5ACD", "#F0F8FF",
"#FFE6F8", "#C71585", "#8B0000"], as_cmap=True)
sns.corrplot(test_data, annot=False, diag_names=False, cmap=cmap)
ax.grid(False)
ax.set_title('Multi-variable correlation significance level')
plt.savefig(model_name+'_multi-variable_correlation', bbox_inches='tight')
# complete coefficient plot - believe this is only for linear regression
sns.coefplot("diagnosis ~ "+' + '.join(features), test_data, intercept=True)
plt.xticks(rotation='vertical')
plt.savefig(model_name+'_coefficient_effects', bbox_inches='tight')
def make_model(model_data_file, feature_list, est):
features = model_data_file[feature_list]
model = est.fit(features, response)
return model
def make_prediction(model_data_file, estimator_list, feature_list, response, feature_values=[]):
# make prediction given feature(s) and model
if not feature_values:
for i in feature_list:
x = float(raw_input("Value for {0}: ".format(i)))
feature_values.append(x)
# go through all the permutations since storing it to csv screwed up the order of all the combinations
for i in itertools.permutations(feature_list):
index = " : ".join(i)
try:
indexer = estimator_list.loc[index]
break
except:
pass
e = indexer.Estimator
parameters = indexer.Accuracy_best
accuracy = indexer.Accuracy
if e == 'linear':
est = LinearRegression(**parameters)
elif e == 'knn':
est = KNeighborsClassifier(**parameters)
elif e == 'logistic':
est = LogisticRegression(**parameters)
elif e == 'gaussian':
est = GaussianNB(**parameters)
elif e == 'svc':
est = SVC(**parameters)
elif e == 'decision_tree':
est = DecisionTreeClassifier(**parameters)
elif e == 'random_forest':
est = RandomForestClassifier(**parameters)
model = make_model(model_data_file, feature_list, est)
prediction = model.predict(feature_values)[0]
print("We are {0}% sure that the your diagnosis for heart disease will be {1}".format(accuracy*100, prediction))
def rfe_trim(features, response, n_features_to_eliminate=4):
# change to a percentage threshold to eliminate
est = LogisticRegression()
selector = RFE(est, n_features_to_select=n_features_to_eliminate)
selector = selector.fit(features, response)
# selected features are assigned rank 1
# print(selector.support_)
# print(selector.ranking_)
trimmings = [i[0] for i in zip(features.columns, selector.support_) if i[1] == False]
trimmed_features = features[trimmings]
print(sorted(zip(selector.ranking_, features.columns)))
return trimmed_features
def tree_trim(features, response, n_features_to_eliminate=4):
# remove features above a certain threnshold
clf = ExtraTreesClassifier()
clf.fit(features, response)
trimmings = [i[1] for i in sorted(zip(clf.feature_importances_, features.columns))][n_features_to_eliminate:]
trimmed_features = features[trimmings]
print([i for i in sorted(zip(clf.feature_importances_, features.columns), reverse=False)])
return trimmed_features
def tree_sort(features, response):
# sorted column names according to most important descending. the higher the number, the more important the feature.
clf = ExtraTreesClassifier()
clf.fit(features, response)
print([i for i in sorted(zip(clf.feature_importances_, features.columns), reverse=True)])
def talking_to_trees(features, response):
mylist = []
limit = 4 # for the number of trees you want to drop
clf = ExtraTreesClassifier()
for i in range(1000):
clf.fit(features, response)
mylist.extend([i[1] for i in sorted(zip(clf.feature_importances_, features.columns),reverse=False)][:limit])
print("Highest scores are the ones that appear most frequenly when asking the tree which {0} are least important.".format(limit))
print(Counter(mylist))
def pretty_print_sorted(dataframe, column=False):
# dataframe.reset_index(inplace=True)
if column: dataframe.sort(column, inplace=True)
pt = PrettyTable()
for i in dataframe.columns:
pt.add_column(i, max_accuracy[i].tolist())
print pt
def save_pickle(data, filename):
with open(filename, 'wb') as f:
pickle.dump(data, f)
# # pickle recovery
# filename = 'recovery'
# pickle = 'combined_df2015-05-06 23:18:19.950979.pickle'
# recover_pickle(pickle, filename)
# get max evaluators
df = load_pickle('stanford_final.pickle')
tuning='Accuracy'
max_accuracy = max_scores(df, tuning)
max_accuracy.to_csv('stanford_max_accuracy.csv')
max_accuracy.to_pickle('stanford_max_accuracy.pickle')
max_accuracy.reset_index(inplace=True)
pretty_print_sorted(max_accuracy, 'Accuracy')
# # comparing trimming methods
# features, response, df = get_sample_dataset()
# rfe_features = rfe_trim(features, response)
# data = (rfe_features, response)
# save_pickle(data, 'rfe_features.csv')
# tree_features = tree_trim(features, response)
# data = (tree_features, response)
# save_pickle(data, 'tree_features.csv')
# # rank important features
# column_labels = ['age', 'sex', 'chest_pain', 'resting_bp', 'cholesterol',
# 'blood_sugar', 'ecg', 'max_hr', 'exercise_induced_angina',
# 'st_depression', 'slope', 'num_major_vessels', 'thal', 'diagnosis']
# response_label = 'diagnosis'
# data_file = 'all_processed_data.csv'
# header = True
# index = True
# features, response, df = data_processor(column_labels, response_label, data_file, header, index)
# talking_to_trees(features, response)
# # make prediction based on best estimator for cleveland data
# f, response, data_file = get_sample_dataset()
# best_estimators = load_pickle('cleveland_max_accuracy.pickle')
# interested_features = ['age', 'resting_bp', 'chest_pain']
# make_prediction(data_file, best_estimators, interested_features, response)
# # make plots
# features, response, df = get_sample_dataset()
# model = LogisticRegression()
# model_name = 'logistic'
# X_train, X_test, y_train, y_test, train_data, test_data = train_test_split(features, response, df)
# make_plot(X_train, y_train, X_test, y_test, test_data, model, model_name, features, 'diagnosis')