handler.py

import os
import pandas as pd
from flask import Flask, jsonify, request
import pickle
from dataprep import TP2EAC 

# load model
model = pickle.load(open('model/modelo.pkl','rb'))

# app
app = Flask(__name__)

# routes
@app.route('/predict', methods=['POST'])

def predict():
    # get data
    data = request.get_json(force=True)

    # converte data into dataframe
    query_df = pd.DataFrame(data)
    query = pd.get_dummies(query_df)

    # instanciar o pre-processamento dos dados
    pipeline = TP2EAC()
    df_data = pipeline.data_preparation(query)
    
    # predictions
    result = model.predict(df_data)
    df_data['prediction'] = result

    # send back to browser
    output = {'results': int(result[0])}

    # return data
    return df_data.to_json( orient = 'records')


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