-
Notifications
You must be signed in to change notification settings - Fork 0
/
analysis_results.py
executable file
·359 lines (283 loc) · 15.4 KB
/
analysis_results.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
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
# -*- coding: utf-8 -*-
"""
Created on Tue Oct 26 14:35:43 2021
@author: padel
"""
#%% Import packages
import pandas as pd
import pickle
import numpy as np
import os
import tqdm as tqdm
from scipy import stats
import json
from statistics import median
from sklearn.metrics import f1_score, confusion_matrix, average_precision_score, roc_auc_score, make_scorer, \
mean_absolute_error, log_loss
from sklearn.preprocessing import LabelEncoder
import shap
from catboost import CatBoostRegressor, CatBoostClassifier, Pool
from catboost.utils import select_threshold
from explainable import (find_explanations_for_completed_cases,
find_explanations_for_running_cases)
from write_results import (prepare_csv_results, write_and_plot_results,
write_scores, write_grid_results, compare_best_validation_curves)
from logme import log_it
from IO import read, write, folders
from os.path import join, exists
import seaborn as sns
import matplotlib.pyplot as plt
np.random.seed(1618)
#%% Implement a prediction
#%% Make the dataframe suitable for the ProM tool
def transform_dataset_for_prom():
ct = 0
X_train['ID ADDED']=np.zeros(len(X_train))
for idx in range(len(X_train)):
X_train['ID ADDED'][idx] = int(ct)
if X_train['ACTIVITY'][idx+1]=='Request created':
ct+=1
X_train.to_csv('traindf.csv', index=False)
#%% It helps me to train, adjust please
def oversampling(X_train, y_train, idxs):
from collections import Counter
counter = Counter(y_train)
print('Before was', counter)
from imblearn.over_sampling import SMOTENC
#Fill idxs with correct number of inputs
sm = SMOTENC(random_state=1618, categorical_features=idxs)
X_train, y_train = sm.fit_resample(X_train, y_train)
counter = Counter(y_train)
print('then is', counter)
return X_train, y_train
def undersampling(X_train, y_train):
raise NotImplementedError()
#Not implemented but already tried
#%% Prepare datasets for running/testing analysis
#%%
# t = 1547221888788
def get_cut_test_set(t=float):
start_end_couple = pickle.load(open('vars/start_end_couple.pkl','rb'))
X_test_idxs = list()
for line in start_end_couple:
if (line[1] < t) and (line[2] > t):
X_test_idxs.append(line[0])
def create_new_test_set(t):
final_testset = pd.DataFrame(columns=X_test.columns)
res_list = list()
ex = 0
for idx in X_test_idxs:
df = non_processed_complete[non_processed_complete['REQUEST_ID']==idx].reset_index(drop=True)
l = 0
try :
while (not ((df['START_DATE'][l] <= t) and (df['END_DATE'][l] >= t))):
l+=1
if l!=0 :
res_list.append([idx, l])
except :
ex += 1
len_test_idxs = {key:item for key,item in zip([i[0] for i in res_list], [i[1] for i in res_list])}
for idx in [i[0] for i in res_list]:
if idx in X_test['REQUEST_ID'].unique():
df = X_test[X_test['REQUEST_ID']==idx].iloc[:len_test_idxs[idx]+1]
final_testset = pd.concat([final_testset, df])
if idx in X_train['REQUEST_ID'].unique():
df = X_train[X_train['REQUEST_ID']==idx].iloc[:len_test_idxs[idx]+1]
final_testset = pd.concat([final_testset, df])
# pickle.dump(final_testset, open('vars/test_set_split_temporarly.pkl','wb'))
# pickle.dump(df_testset, open('vars/test_set_split_temporarly_with_y.pkl','wb'))
def fill_y_values():
df_testset = pickle.load(open('vars/test_set_split_temporarly.pkl','rb')).reset_index(drop=True)
df_testset['lead_time'] = np.zeros(len(df_testset))
y_dict = dict()
for idx in df_total['REQUEST_ID'].unique():
df = df_total[df_total['REQUEST_ID']==idx]
if len(set(df['lead_time']))!=1:
print(idx)
else :
y_dict[str(idx)] = list(set(df['lead_time']))[0]
l = list()
for idx in df_testset.index:
l.append(y_dict[str(df_testset.loc[idx]['REQUEST_ID'])])
df_testset['lead_time'] = l
#%%
def find_sub_list(sl,l):
results=[]
sll=len(sl)
for ind in (i for i,e in enumerate(l) if e==sl[0]):
if l[ind:ind+sll]==sl:
results.append((ind,ind+sll-1))
return results
#%%
# ratio_cut = 5
# maximize = False
# is_real = False
# predict_activities = 'Pending Liquidation Request'
# experiment, maximize, predict_activities = 'indipendent_activity', False, 'Pending Liquidation Request'
# maximize = 'max'*maximize + 'min'*(not maximize)
results = pickle.load(open('/home/padela/Scrivania/PhD/Prescriptive-Analytics/vars/lead_time/test_results_min_None.pkl', 'rb'))
def results_to_df(ratio_cut=float, maximize=bool, is_real=True, predict_activities=str, filling=False, save=False):
print(ratio_cut, maximize, is_real, predict_activities)
#Set values for importer
maximize = 'max'*maximize + 'min'*(not maximize)
is_real = '_real'*(is_real)
ratio_cut = float(ratio_cut)
#Import results
# results = pickle.load(open(f'vars/{experiment}/test_results_ratiocut{ratio_cut}_{maximize}_{predict_activities}{is_real}.pkl','rb'))
results = pickle.load(open(f'vars/{experiment}/test_results_{maximize}_{predict_activities}.pkl', 'rb'))
#Drop lines in which there's not a result
results = [res for res in results if not isinstance(res[-1], str)]
#Create and fill a dataset
df_res = pd.DataFrame()
df_res['result'] = [res[0] for res in results]
df_res['len_trace_after_cut'] = [res[1] for res in results]
df_res['#posssible_next_activity'] = [res[2] for res in results]
df_res['score_reality'] = [res[3] for res in results]
df_res['score_recommendation'] = [res[4] for res in results]
df_res['history'] = [res[5][:-1] for res in results]
df_res['suggested_activity'] = [res[6] for res in results]
#Create artificial columns
df_res['ratio_recommendation/reality'] = 1-(np.array(df_res['score_recommendation']))/(np.array(df_res['score_reality']))
df_res['diff_reality_recommendation'] = np.array(df_res['score_reality'] - np.array(df_res['score_recommendation']))
df_res['original_trace_kpi'] = [((res[5][int(res[2]):]).count(predict_activities)>0)*1 for res in results]
df_res['my_score'] = 1- ((np.array(df_res['score_recommendation'])) / (np.array(df_res['score_reality'])))
abs_score = round(df_res['result'].mean()*100, 2)
poss_acts = df_res['#posssible_next_activity'].mean()
avg_ratio = round(df_res['ratio_reality_recommendation'].mean(), 2)
print(f'For cut {ratio_cut}:')
print(f'The recommendation is correct in the {abs_score} % of cases')
print(f'The The average ratio between score of real continuation and the recommended continuation is {avg_ratio}')
print(f'The average number of actions possible is {poss_acts}\n')
if save :
pickle.dump(df_res, open(f'vars/{experiment}/df_res{ratio_cut}_{maximize}_{predict_activities}{is_real}.pkl','wb'))
return df_res
def plot_qini_curve_comparison(maximize=False, is_real=True, predict_activities=str):
#VA BENE COME È SCRITTO SOLO PER MAXIMIZE = FALSE
curve_val_all = list()
curve_val_real = list()
random_class = list()
maximize = 'max'*maximize + 'min'*(not maximize)
is_real = '_real'*(is_real)
for ratio_cut in range(4,10):
ratio_cut = float(ratio_cut)
df = pickle.load(open(f'vars/{experiment}/df_res{ratio_cut}_{maximize}_{predict_activities}{is_real}.pkl','rb'))
curve_val_all.append(df['result'].mean()*100)
curve_val_real.append((sum(df['score_reality']>=df['score_recommendation'])/len(df))*100)
random_class.append(100/(df['#posssible_next_activity'].mean()))
#Plot the curve
sns.set_style('darkgrid')
plt.title(label=f'Qini curve for minimization of {predict_activities} \n percent of improvements over all dataframe')
plt.xlabel('Percentage of cut on the total length of trace')
plt.ylabel('Score')
sns.lineplot([ratio_cut*10 for ratio_cut in range(4,10)], curve_val_all, color='darkred')
sns.lineplot([ratio_cut*10 for ratio_cut in range(4,10)], curve_val_real, color='blue')
sns.lineplot([ratio_cut*10 for ratio_cut in range(4,10)], random_class, color='black', linestyle='--')
plt.legend(['accuracy compared with all possible ends', 'accuracy compared with real ends', '1/possible_activities'])
def plot_qini_curve_score(maximize=False, is_real=True, predict_activities=str):
#VA BENE COME È SCRITTO SOLO PER MAXIMIZE = FALSE
percentage = list()
random_class = list()
maximize = 'max'*maximize + 'min'*(not maximize)
is_real = '_real'*(is_real)
for ratio_cut in range(4,10):
ratio_cut = float(ratio_cut)
df = pickle.load(open(f'vars/{experiment}/df_res{ratio_cut}_{maximize}_{predict_activities}{is_real}.pkl','rb'))
random_class.append((1/(df['#posssible_next_activity'].mean()))*100)
percentage.append((df['ratio_reality_recommendation'].mean()-1)*100)
sns.set_style('darkgrid')
plt.title(label=f'Qini curve for minimization of {predict_activities} \n average of improvement')
plt.xlabel('Percentage of cut on the total length of trace')
plt.ylabel('Score')
sns.lineplot([ratio_cut*10 for ratio_cut in range(4,10)], random_class, color='black', linestyle='--')
sns.lineplot([ratio_cut*10 for ratio_cut in range(4,10)], percentage, color='darkgreen')
plt.legend(['1/possible_activities','Percent of improvement compared to reality'])
def generate_table_values(maximize=False, is_real=True, predict_activities=str):
tables = dict()
maximize = 'max'*maximize + 'min'*(not maximize)
is_real = '_real'*(is_real)
for ratio_cut in range(4,10):
ratio_cut = float(ratio_cut)
df = pickle.load(open(f'vars/{experiment}/df_res{ratio_cut}_{maximize}_{predict_activities}{is_real}.pkl','rb'))
d = len(df)
kpi_positive = df[df['original_trace_kpi']>0]
kpi_positive_ = sum((kpi_positive['score_reality'] - kpi_positive['score_recommendation']).values>=0)
kpi_null = df[df['original_trace_kpi']==0]
kpi_null_ = sum((kpi_null['score_reality'] - kpi_null['score_recommendation'])>=0)
tables[str(ratio_cut)] = np.array([[kpi_null_, len(kpi_null)-kpi_null_, round(kpi_null_/d, 3), round((len(kpi_null)-kpi_null_)/d, 3)],
[kpi_positive_, len(kpi_positive)-kpi_positive_, round(kpi_positive_/d, 3), round((len(kpi_positive)-kpi_positive_)/d, 3)]])
return tables
def target_val_distribution_plot(ratio_cut=int, bw_method= None, maximize=False, is_real=True, predict_activities=str):
ratio_cut = float(ratio_cut)
maximize = 'max'*maximize + 'min'*(not maximize)
is_real = '_real'*(is_real)
df = pickle.load(open(f'vars/{experiment}/df_res{ratio_cut}_{maximize}_{predict_activities}{is_real}.pkl','rb'))
df = df_res[df_res['#posssible_next_activity']>1]
# df = df[df['diff_reality_recommendation']!=0]
sns.set_style('darkgrid')
fig, ax = plt.subplots(1,2)
# plt.suptitle(t = predict_activities+' Distribution of target values without considering 0-values')
if predict_activities == 'Back-Office Adjustment Requested':
ax[0].set_xlim(-0.05,.08)
ax[1].set_ylim(-0.01, 0.05)
# ax[0].set_xlim(-1e7, 1e7)
# ax[1].set_ylim(0e7, 1e7)
sns.kdeplot(data = df['score_recommendation'], ax = ax[0])
sns.kdeplot(data = df['score_reality'], ax = ax[0])
sns.kdeplot(data = df['score_recommendation'] - df['score_reality'] , ax = ax[0])
ax[0].legend(['score_recommendation', 'score_reality' ,'recommendation - reality '])
sns.boxplot(data = df[['score_recommendation', 'score_reality']], orient="l", palette="Set1", ax = ax[1], showmeans=True)#.xlabel(' ')
sns.boxplot(data = df['ratio_recommendation/reality'], orient="l", palette="Set1", showmeans=True)#.xlabel(' ')
sns.kdeplot(data = df['ratio_recommendation/reality'])#.xlabel(' ')
plt.ylim(-1,1)
plt.title('Improvement on time evaluated using the formula below')
print('Means are ', np.mean(df['score_recommendation']),' ', np.mean(df['score_reality']))
print('Stds are ', np.std(df['score_recommendation']),' ', np.std(df['score_reality']))
print(stats.describe(np.array(df['score_recommendation'])))
def check_equal_values(ratio_cut=int, bw_method= None, maximize=False, is_real=True, predict_activities=str):
#Check the effectiveness on the 0-diff values
ratio_cut = float(ratio_cut)
maximize = 'max'*maximize + 'min'*(not maximize)
is_real = '_real'*(is_real)
df = pickle.load(open(f'vars/{experiment}/df_res{ratio_cut}_{maximize}_{predict_activities}{is_real}.pkl','rb'))
#Filter on 0-diff values
sns.set_style('darkgrid')
plt.ylim(-0.001, 0.04)
df = df[df['diff_reality_recommendation']==0]
plt.title(f'Distribution of original target values without differences in with/without \n reccomendation for {predict_activities}')
# plt.ylim(0,0.06)
plt.boxplot(df['score_reality'], showmeans=True)
#%% Run cases
if __name__ == "__main__":
maximize = False
is_real = False
predict_activities = 'Back-Office Adjustment Requested' # 'Pending Liquidation Request' #
print(f'Remember to set parameters in code, now they\' re : \nmaximize = {maximize} \nis_real {is_real} \nactivity: {predict_activities}')
for ratio_cut in range(4,10):
df_res = results_to_df(ratio_cut=ratio_cut, maximize=maximize, is_real=is_real, predict_activities=predict_activities, save=True)
tables = generate_table_values(maximize=maximize, is_real=is_real, predict_activities=predict_activities)
# plot_qini_curve_score(maximize=maximize, is_real=is_real, predict_activities=predict_activities)
plot_qini_curve_comparison(maximize=maximize, is_real=is_real, predict_activities=predict_activities)
target_val_distribution_plot(ratio_cut=5, bw_method=0.15, maximize=maximize, is_real=is_real, predict_activities=predict_activities)
check_equal_values(ratio_cut = 5, bw_method = .15, maximize=False, is_real=is_real, predict_activities=predict_activities)
#%%
'''
l = len(X_test.REQUEST_ID.unique())
l_n = len(df_test.REQUEST_ID.unique())
for idx in df_test.REQUEST_ID.unique():
df = Counter(list(df_test[df_test['REQUEST_ID']==idx]['ACTIVITY']))
if df['Back-Office Adjustment Requested']!=0:
print(df['Back-Office Adjustment Requested']/l_n)
'''
# for idx in X_test['REQUEST_ID'].unique():
# df = X_test[X_test['REQUEST_ID']==idx]
# if len(set(df['y']))>1:
# print(idx)
# problema è su real, backoffice, ratio_cut4
#%% THey were for lead time analysis
'''
results = pickle.load(open('vars/lead_time/test_results_min_None.pkl','rb'))
df_res[df_res['diff_reality_recommendation']!=0]['diff_reality_recommendation'].plot.kde()
np.std(df_res[df_res['diff_reality_recommendation']!=0]['diff_reality_recommendation'])
np.std(df_res['diff_reality_recommendation'])
sns.boxplot(df_res['diff_reality_recommendation'])
'''