-
Notifications
You must be signed in to change notification settings - Fork 0
/
app.py
186 lines (153 loc) · 8.25 KB
/
app.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
import traceback
from flask import render_template, request, redirect, url_for
import flask
import logging.config
import src.config as config
from src.helpers.helpers import read_csv_from_s3,load_saved_model
# from app.models import Tracks
from flask import Flask
import pickle
import boto3
import argparse
from io import BytesIO
import pandas as pd
from sklearn.preprocessing import LabelEncoder
import numpy as np
from flask_sqlalchemy import SQLAlchemy
import os
from src.data_ingest_schema_create import UserInput
# Initialize the Flask application
app = Flask(__name__,template_folder='app/templates/')
app.config.from_pyfile(os.path.join('config','flask_config.py'))
# Initialize the database
db = SQLAlchemy(app)
logger = logging.getLogger(__name__)
# Use pickle to load in the pre-trained modeldd
s3 = boto3.resource('s3')
parser = argparse.ArgumentParser()
parser.add_argument("--bucket_name", default= config.DEFAULT_BUCKET_NAME, help="S3 bucket to upload the source data to. Default:nw-shreyassabnis-msia423")
args = parser.parse_args()
df_train = read_csv_from_s3(args.bucket_name, config.DEFAULT_BUCKET_FOLDER, config.TRAIN_DATA)
genders = list(df_train.gender.unique())
signup_methods = list(df_train.signup_method.unique())
languages = list(df_train.language.unique())
affiliate_channels = list(df_train.affiliate_channel.unique())
country_map = read_csv_from_s3(args.bucket_name, config.DEFAULT_BUCKET_FOLDER, config.COUNTRY_MAP)
feature_mode = read_csv_from_s3(args.bucket_name, config.FEATURE_FOLDER, config.MODE_FEATURES_FILE_NAME)
training_data = read_csv_from_s3(args.bucket_name, config.FEATURE_FOLDER, config.TRAIN_FEATURE_FILE)
feature_df = read_csv_from_s3(args.bucket_name, config.FEATURE_FOLDER, config.FEATURE_FILE_NAME)
train_users = read_csv_from_s3(args.bucket_name, config.DEFAULT_BUCKET_FOLDER, config.TRAIN_DATA)
labels = read_csv_from_s3(args.bucket_name, config.FEATURE_FOLDER, config.LABEL_FILE_NAME)
le = LabelEncoder()
y = le.fit_transform(labels)
user_list = list(feature_df.head(config.NUM_USER_ID_TO_DISPLAY)['userid'])
def get_country_name(country_map, country):
return country_map.loc[country_map.Code == country,'Name'].values[0]
s3 = boto3.resource('s3')
try:
model = load_saved_model(args.bucket_name, config.MODEL_S3_LOCATION)
logger.info("Loaded saved model %s from bucket %s",config.MODEL_S3_LOCATION , args.bucket_name)
except FileNotFoundError as e:
logger.error('No saved model found! %s',e)
@app.route('/', methods=['GET','POST'])
def index():
"""Main view that lists songs in the database.
Create view into index page that uses data queried from Track database and
inserts it into the msiapp/templates/index.html template.
Returns: rendered html template
"""
if flask.request.method == 'GET':
return(flask.render_template('index.html', genders = genders, signup_methods = signup_methods, languages = languages, affiliate_channels = affiliate_channels ))
if flask.request.method == 'POST':
gender_resp = flask.request.form['gender']
signupmethod_resp = flask.request.form['signupmethod']
language_resp = flask.request.form['language']
affiliatechannel_resp = flask.request.form['affiliatechannel']
gender = 'gender_'+gender_resp
signupmethod = 'signup_method_'+signupmethod_resp
language = 'language_'+language_resp
affiliatechannel = 'affiliate_channel_' +affiliatechannel_resp
input_variables = pd.DataFrame([[gender, signupmethod, language, affiliatechannel]],columns=['gender', 'signupmethod', 'language', 'affiliatechannel'])
upd_col_lst = [gender, signupmethod, language, affiliatechannel]
for upd_col in upd_col_lst:
try:
feature_mode[upd_col]=1
except:
logger.warning("column %s does not exist in training data, ignoring update", upd_col)
prediction = model.predict_proba(feature_mode[list(training_data)].values)
y_pred_names = []
y_pred_prob = []
for i in range(len(prediction)):
y_pred_names.append(list(le.inverse_transform(np.argsort(prediction[i])[::-1])[:2]))
a =prediction[i].tolist()
a.sort(reverse=True)
a=a[:2]
a = [round(x,2) for x in a]
y_pred_prob.append(a)
pred_val_tbl = [get_country_name(country_map,y_pred_names[0][0]),get_country_name(country_map,y_pred_names[0][1])]
pred_val = y_pred_names[0][0]+","+y_pred_names[0][1]
if(app.config['RDS_FLAG']=="T"):
logger.debug("Adding new record to database in AWS RDS.")
else:
logger.debug("Adding new record to the local SQLite database.")
new_entry = UserInput(
Gender = gender_resp
,SignupMethod = signupmethod_resp
,Language = language_resp
,AffiliateChannel = affiliatechannel_resp)
db.session.add(new_entry)
db.session.commit()
logger.debug("New record added to the database")
return flask.render_template('index.html',
original_input={'Gender':gender_resp,
'signupmethod':signupmethod_resp,
'language':language_resp,
'affiliatechannel':affiliatechannel_resp
},
result = pred_val, result_tbl=pred_val_tbl, probs=y_pred_prob[0] , genders = genders, signup_methods = signup_methods, languages = languages, affiliate_channels = affiliate_channels
)
@app.route('/userid', methods=['GET','POST'])
def userid():
"""Main view that lists songs in the database.
Create view into index page that uses data queried from Track database and
inserts it into the msiapp/templates/index.html template.
Returns: rendered html template
"""
if flask.request.method == 'GET':
return(flask.render_template('userid.html',user_lst = user_list))
if flask.request.method == 'POST':
userid = flask.request.form['userid']
#print(userid)
data_point = feature_df.loc[feature_df.userid== userid, list(training_data)]
#print(data_point)
prediction = model.predict_proba(data_point.values)
#print('prediction ',prediction)
y_pred_names = []
y_pred_prob = []
for i in range(len(prediction)):
y_pred_names.append(list(le.inverse_transform(np.argsort(prediction[i])[::-1])[:2]))
#y_pred_prob.append(list(np.argsort(prediction[i])[::-1])[:2])
a =prediction[i].tolist()
a.sort(reverse=True)
a=a[:2]
a = [round(x,2) for x in a]
y_pred_prob.append(a)
if len(y_pred_names)>0:
gender_resp = train_users.loc[train_users.id==userid,'gender'].values[0]
signupmethod_resp = train_users.loc[train_users.id==userid,'signup_method'].values[0]
language_resp = train_users.loc[train_users.id==userid,'language'].values[0]
affiliate_channel_resp = train_users.loc[train_users.id==userid,'affiliate_channel'].values[0]
pred_val_tbl = [get_country_name(country_map,y_pred_names[0][0]),get_country_name(country_map,y_pred_names[0][1])]
pred_val = y_pred_names[0][0]+","+y_pred_names[0][1]
return flask.render_template('userid.html',
original_input={'Gender':gender_resp,
'signupmethod':signupmethod_resp,
'language':language_resp,
'affiliatechannel':affiliate_channel_resp
},
result = pred_val ,result_tbl=pred_val_tbl, probs=y_pred_prob[0], user_lst = user_list
)
else:
return flask.render_template('error.html')
if __name__ == '__main__':
app.run(host = app.config['HOST'],port=app.config['PORT'],debug=app.config['DEBUG'])