app.py

import gradio as gr
from PIL import Image
import requests
import hopsworks
import joblib
import pandas as pd
import numpy as np

project = hopsworks.login(project='suyiw000')

fs = project.get_feature_store()


mf = project.get_model_registry()
model = mf.get_model("food_model", version=1)
model_dir = model.download()
model = joblib.load(model_dir + "/food_model.pkl")
print("Model downloaded")


market = ['Badakhshan', 'Badghis', 'Baghlan', 'Balkh', 'Bamyan', 'Daykundi', 'Farah', 'Faryab', 'Ghazni', 'Ghor', 'Hilmand', 'Hirat', 'Jawzjan' 'Kabul', 'Kandahar', 'Kapisa', 'Khost', 'Kunar', 'Kunduz', 'Laghman', 'Logar', 'Maidan Wardak', 'Nangarhar', 'Nimroz', 'Nuristan', 'Paktika', 'Paktya', 'Panjsher', 'Parwan', 'Samangan', 'Sar-e-Pul', 'Takhar', 'Uruzgan', 'Zabul']
commodity = ['Bread', 'Oil_cooking', 'Pulses', 'Rice_high', 'Rice_low', 'Salt', 'Sugar', 'Wheat', 'Wheatflour_high', 'Wheatflour_low']


def predict_price(year, month, market, food):
  market_empty = np.zeros(34)
  market_name = ['Badakhshan', 'Badghis', 'Baghlan', 'Balkh', 'Bamyan', 'Daykundi', 'Farah', 'Faryab', 'Ghazni', 'Ghor', 'Hilmand', 'Hirat', 'Jawzjan' 'Kabul', 'Kandahar', 'Kapisa', 'Khost', 'Kunar', 'Kunduz', 'Laghman', 'Logar', 'Maidan Wardak', 'Nangarhar', 'Nimroz', 'Nuristan', 'Paktika', 'Paktya', 'Panjsher', 'Parwan', 'Samangan', 'Sar-e-Pul', 'Takhar', 'Uruzgan', 'Zabul']
  market = []
  for i in range(34):
    temp_market = market_empty.copy()
    temp_market[i] = 1.0
    market.append(temp_market)

  commodity_empty = np.zeros(10)
  commodity_name = ['Bread', 'Oil_cooking', 'Pulses', 'Rice_high', 'Rice_low', 'Salt', 'Sugar', 'Wheat', 'Wheatflour_high', 'Wheatflour_low']
  commodity=[]
  for i in range(10):
    commodity_array = commodity_empty.copy()
    commodity_array[i] = 1.0
    commodity.append(commodity_array)

  commodity_with_names = dict(zip(commodity_name, commodity))
  arrays_with_names = dict(zip(market_name, market))

  date = ((year*10000+month*100+15)-20200000)/100000

  input_data = np.concatenate([arrays_with_names[market], commodity_with_names[food], [date]]).reshape(1, -1)
 
  prediction = model.predict(input_data)

  return prediction


demo = gr.Interface(
    fn = predict_price,
    title = "AFG FOOD PRICE PREDICTION",
    allow_flagging="never",
    inputs=[
        gr.Number(label="Year"),
        gr.Number(label="Mouth"),
        gr.Dropdown(choices=market, label="Market"),
        gr.Dropdown(choices=commodity, label="Food Type")
        ],
    outputs="text"
)
demo.launch(debug=True)