streamlit_app.py

import streamlit as st
from datetime import datetime
import pandas as pd
import joblib
import herepy
import os
from dotenv import load_dotenv

dotenv_path = os.path.join(os.path.dirname(__file__), '.env')
load_dotenv(dotenv_path)
HERE_API_KEY = os.getenv('HERE_API_KEY')

X = ['pickup_datetime',
        'pickup_longitude',
        'pickup_latitude',
        'dropoff_longitude',
        'dropoff_latitude']

def geocoder(address, key=HERE_API_KEY):
    geoapi = herepy.GeocoderApi(api_key=key)
    result = geoapi.free_form(address).as_dict()
    coords = result["items"][0]["position"]
    coords = {k.lower().replace('lng', 'lon'):v for k, v in coords.items()}
    return coords

def format_input(pickup, dropoff):
    pickup_datetime = datetime.utcnow()
    formated_input = {'pickup_datetime': str(pickup_datetime)+' UTC',
                        'pickup_latitude': float(pickup['lat']),
                        'pickup_longitude': float(pickup['lon']),
                        'dropoff_latitude': float(dropoff['lat']),
                        'dropoff_longitude': float(dropoff['lon'])}
    return formated_input

def main():
    st.set_page_config(page_title="nyc-taxifare-predictor",
                    page_icon=":oncoming_taxi:",
                    layout="centered")
    st.markdown('https://github.com/Guli-Y/NYCtaxifarePredictor')
    pipe = joblib.load('model.joblib')
    print('------------ loaded model ---------------')
    st.header('NYC Taxi Fare Predictor :taxi:')
    st.write('Please type in pickup and dropoff locations to get predicted taxi fare amount!')
    # input
    pickup_address = st.text_input('pickup address', '45 Rockefeller Plaza, New York, NY 10111')
    dropoff_address = st.text_input('dropoff address', '334 Furman St, Brooklyn, NY 11201')
    # get coords
    pickup_coords = geocoder(pickup_address)
    dropoff_coords = geocoder(dropoff_address)
    # input dictionary
    df = pd.DataFrame([format_input(pickup_coords, dropoff_coords)])
    df = df[X]
    result = pipe.predict(df)
    fare = round(float(result[0]), 3)
    st.write(':oncoming_taxi: *Fare Amount*', fare, ':heavy_dollar_sign:')
    locations = pd.DataFrame([pickup_coords, dropoff_coords])
    st.map(data=locations, zoom=11)
    st.markdown('Note: The model was trained on the data sets that were recorded between 2008 and 2015, and thus the predicted fare might not apply to current dates. This project is purely for training purpose.')

if __name__ == '__main__':
    main()