forked from leanderme/sytora
-
-
Notifications
You must be signed in to change notification settings - Fork 4
/
train.py
146 lines (95 loc) · 3.81 KB
/
train.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
import pandas as pd
import numpy as np
import re
import os
import glob
import pandas as pd
import matplotlib.pyplot as plt
from sklearn.naive_bayes import MultinomialNB
from sklearn.cross_validation import train_test_split
from sklearn.externals import joblib
result = pd.read_csv('./disease-symptom-db.csv', encoding='utf-8', index_col=None, header=0)
def isValid(cui):
cui = str(cui)
pattern = re.compile("C\\d{7}")
if not pattern.match(cui):
return False
return True
def cuiToNumber(cui):
return cui.strip("C").strip("0")
def convertCUI(cui):
cui = str(cui)
if not isValid(cui):
return "C" + cui.zfill(7)
else:
return cui
def clean(the_string):
return str(the_string.encode('utf-8'))
result['Disease'] = result['Disease'].apply(convertCUI)
result['Symptom'] = result['Symptom'].apply(convertCUI)
result.to_csv("./disease-symptom-db.csv",index=False)
df_foreign = pd.read_csv('./DiseaseSymptomKB.csv', encoding='utf-8', index_col=None, header=0)
result = result.append(df_foreign)
result.to_csv("./disease-symptom-merged.csv",index=False)
result['Disease'] = result['Disease'].astype(str)
result['Symptom'] = result['Symptom'].astype(str)
result['Symptom'].replace('', np.nan, inplace=True)
result.dropna(subset=['Symptom'], inplace=True)
result['Disease'].replace('', np.nan, inplace=True)
result.dropna(subset=['Disease'], inplace=True)
df = pd.DataFrame(result)
df.columns
df_1 = pd.get_dummies(df.Symptom)
df_s = df['Disease']
df_pivoted = pd.concat([df_s,df_1], axis=1)
df_pivoted.drop_duplicates(keep='first',inplace=True)
cols = df_pivoted.columns
cols = cols[1:] # skip 'Disease'
df_pivoted = df_pivoted.groupby('Disease').sum()
df_pivoted = df_pivoted.reset_index()
all_files_ml = "./data/all-files-for-ml"
df_pivoted.to_csv(os.path.join(all_files_ml, "all_pivoted.csv"), index=False)
cols = df_pivoted.columns
cols = cols[1:] # skip 'title'
x = df_pivoted[cols] # symptom rows
y = df_pivoted['Disease'] # diseases
x.to_csv(os.path.join(all_files_ml, "all_x.csv"), index=False)
x_train, x_test, y_train, y_test = train_test_split(x, y, test_size=0.33, random_state=42)
mnb = MultinomialNB()
mnb = mnb.fit(x_train, y_train)
mnb.score(x_test, y_test)
mnb_tot = MultinomialNB()
mnb_tot = mnb_tot.fit(x, y)
mnb_tot.score(x, y)
disease_pred = mnb_tot.predict(x)
disease_real = y.values
for i in range(0, len(disease_real)):
if disease_pred[i]!=disease_real[i]:
print ('Pred: {0} Actual:{1}'.format(disease_pred[i], disease_real[i]))
joblib.dump(mnb, os.path.join(all_files_ml, 'all_mnb.pkl'), protocol=2)
data = pd.read_csv(os.path.join(all_files_ml, "all_x.csv"))
df = pd.DataFrame(data)
cols = df.columns
features = cols # = symptoms
features_raw = [str(features[x]) for x in range(len(features))]
features_raw = ','.join(map(str, features_raw))
# convert feature array into dict of symptom: index
feature_dict = {}
for i,f in enumerate(features):
feature_dict[f] = i
def findFeatures(disease):
return result.loc[result['Disease'] == disease]["Symptom"].values.astype(str)
sample = np.zeros((len(features),), dtype=np.int)
sampe = sample.tolist()
search = ["C0857794", "C0149793", "C0000786"]
for i,s in enumerate(search):
sample[feature_dict[s]] = 1
sample = np.array(sample).reshape(1,len(sample))
results = mnb.predict_proba(sample)[0]
# gets a dictionary of {'class_name': probability}
prob_per_class_dictionary = dict(zip(mnb.classes_, results))
# gets a list of ['most_probable_class', 'second_most_probable_class', ..., 'least_class']
results_ordered_by_probability = map(lambda x: {"disease": x[0],"prop": x[1] * 100, "sy": findFeatures(x[0])}, sorted(zip(mnb.classes_, results), key=lambda x: x[1], reverse=True))
print (list(results_ordered_by_probability))
#store the predicted probabilities for class 1
y_pred_prob = mnb.predict_proba(sample)[0]