app.py

# Flask
from flask import Flask, render_template, request, redirect, url_for, session, abort, make_response

# MongoDB and Sessions
from flask.ext.session import Session
from pymongo import MongoClient

#ML Libs
import numpy as np
from sklearn.svm import SVC
from sklearn.externals import joblib
from sklearn.preprocessing import scale

# Miscellaneous
import os, logging, json, datetime

tmpl_dir = os.path.join(os.path.dirname(os.path.abspath(__file__)), 'templates')
app = Flask(__name__)

# MongoDB Setup
client = MongoClient(os.getenv('MONGOHQ_URL'))
db = client.bc15

# MongoDB Session Setup
SESSION_TYPE = "mongodb"
SESSION_MONGODB = client
SESSION_MONGODB_DB = "bc15"
SESSION_MONGODB_COLLECT = "sessions"
app.secret_key = '\xcbC\xb0\x0b\xe6\xe7\xc2\xd3u\xbf\xff\x9b\xfb\xb8\xb1\xb9^Y\xdbQ\xba\x7f\xabl'

app.config.from_object(__name__)
Session(app)

@app.before_first_request
def setup_logging():
	if not app.debug:
		# In production mode, add log handler to sys.stderr.
		app.logger.addHandler(logging.StreamHandler())
		app.logger.setLevel(logging.INFO)

@app.route('/')
def index():
	return render_template('index.html', template_folder=tmpl_dir)

@app.route('/dashboard')
def dashboard():
	return render_template('dashboard.html', template_folder=tmpl_dir)

@app.route('/history')
def history():
	_id = "session:" + str(session.sid)
	data = db.history.find_one({ "id" : _id })
	if data != None:
		return render_template('history.html', data=data, template_folder=tmpl_dir)
	return render_template('error.html', template_folder=tmpl_dir, error_msg="No Search History Found", 
		return_home="You have to search the map before you can see your search history."	
	)

@app.route('/clear')
def clear_history():
	_id = "session:" + str(session.sid)
	data = db.history.remove({ "id" : _id })
	return redirect(url_for('history'))


@app.route('/predict', methods=["POST"])
def predict():
	_id = "session:" + str(session.sid)
	hours, minutes = [int(i) for i in request.form["timepicker"].split(':')]
	time = (60 * hours) + minutes
	svm = get_predictor()
	if svm != None:
		data = scale(np.array([[float(time), float(request.form["MapLat"]), float(request.form["MapLon"])]]))
		prediction = svm.predict(data)
	probability = round(np.amax(svm.predict_proba(data)) * 100, 2)
	db.history.update({ "id" : _id }, {"$push": { "history": {
		"timeOfDay": int(time),
		"latitude": float(request.form["MapLat"]),
		"longitude": float(request.form["MapLon"]),
		"address": str(request.form["gmaps-input"]),
		"probability": float(probability),
		"prediction": str(prediction[0])[2:-1]
	}}}, upsert=True);
	return render_template('dashboard.html', prediction=str(prediction[0])[2:-1], addr=request.form["gmaps-input"], lat=request.form["MapLat"], 
		lng=request.form["MapLon"], probability=probability, template_folder=tmpl_dir)

@app.errorhandler(401)
def unauthorized(error):
	return render_template('error.html', template_folder=tmpl_dir, error=401, error_msg="Unauthorized", 
		return_home="You must be logged in to access this page!"	
	)

@app.errorhandler(500)
def internal_server(e):
	return render_template('error.html', template_folder=tmpl_dir, error=500, error_msg="Internal Server Error", 
		return_home="Something went wrong! Let us know if it happens again!"	
	)

@app.errorhandler(404)
def page_not_found(e):
	return render_template('error.html', template_folder=tmpl_dir, error=404, error_msg="Page Not Found",
		return_home="We can't find what you're looking for."
	)

def get_predictor():
	try:
		svm = joblib.load("data/svm.pkl")
	except:
		crime_data = scale(np.loadtxt(open("data/crimeData.csv", "rb"), delimiter=",", skiprows=1))
		crime_target = np.loadtxt(open("data/crimeLabels.csv", "rb"), delimiter=",", dtype=str)
		svm = SVC(kernel="rbf", probability=True).fit(crime_data, crime_target)
		joblib.dump(svm, "data/svm.pkl")
	finally:
		return svm

if __name__ == "__main__":
	port = int(os.environ.get("PORT", 5000))
	app.run(host='0.0.0.0', port=port)