Agrego ejemplo de crear api Flask para machine learning

api_ml/api_train_model.py

@@ -0,0 +1,59 @@
+import pandas as pd
+import numpy as np
+from sklearn.preprocessing import MinMaxScaler
+
+from utiles import *
+
+df = pd.read_csv('time_series.csv',  parse_dates=[0], header=None,index_col=0, names=['fecha','unidades'])
+df['weekday']=[x.weekday() for x in df.index]
+df['month']=[x.month for x in df.index]
+print(df.head())
+
+EPOCHS=40
+PASOS=7
+
+scaler = MinMaxScaler(feature_range=(-1, 1))
+
+reframed = transformar(df, scaler)
+
+reordenado=reframed[ ['weekday','month','var1(t-7)','var1(t-6)','var1(t-5)','var1(t-4)','var1(t-3)','var1(t-2)','var1(t-1)','var1(t)'] ]
+reordenado.dropna(inplace=True)
+
+training_data = reordenado.drop('var1(t)', axis=1)
+target_data=reordenado['var1(t)']
+cant = len(df.index)
+valid_data = training_data[cant-30:cant]
+valid_target=target_data[cant-30:cant]
+
+training_data = training_data[0:cant]
+target_data=target_data[0:cant]
+print(training_data.shape, target_data.shape, valid_data.shape, valid_target.shape)
+print(training_data.head())
+
+model = crear_modeloEmbeddings()
+
+continuas = training_data[['var1(t-7)','var1(t-6)','var1(t-5)','var1(t-4)','var1(t-3)','var1(t-2)','var1(t-1)']]
+valid_continuas = valid_data[['var1(t-7)','var1(t-6)','var1(t-5)','var1(t-4)','var1(t-3)','var1(t-2)','var1(t-1)']]
+
+history = model.fit([training_data['weekday'],training_data['month'],continuas], target_data, epochs=EPOCHS,
+                 validation_data=([valid_data['weekday'],valid_data['month'],valid_continuas],valid_target))
+
+results = model.predict([valid_data['weekday'],valid_data['month'],valid_continuas])
+
+print( 'Resultados escalados',results )
+inverted = scaler.inverse_transform(results)
+print( 'Resultados',inverted )
+
+save_object('scaler_time_series.pkl', scaler)
+model.save('red_time_series.h5')
+model.save_weights("pesos.h5")
+
+#loaded_model = load_model('red_time_series.h5')
+loaded_model = crear_modeloEmbeddings()
+loaded_model.load_weights("pesos.h5")
+
+results = loaded_model.predict([valid_data['weekday'],valid_data['month'],valid_continuas])
+print( 'Resultados escalados',results )
+loaded_scaler = load_object('scaler_time_series.pkl')
+inverted = loaded_scaler.inverse_transform(results)
+print( 'Resultados',inverted )

api_ml/server.py

@@ -0,0 +1,46 @@
+"""Filename: server.py
+"""
+import pandas as pd
+from sklearn.externals import joblib
+from flask import Flask, jsonify, request
+
+from utiles import *
+
+app = Flask(__name__)
+
+@app.route('/predict', methods=['POST'])
+def predict():
+	"""API request
+	"""
+	try:
+		req_json = request.get_json()
+		input = pd.read_json(req_json, orient='records')
+	except Exception as e:
+		raise e
+
+	if input.empty:
+		return(bad_request())
+	else:
+		#Load the saved model
+		print("Cargar el modelo...")
+		loaded_model = crear_modeloEmbeddings()
+		loaded_model.load_weights("pesos.h5")
+
+		print("Hacer Pronosticos")
+		continuas = input[['var1(t-7)','var1(t-6)','var1(t-5)','var1(t-4)','var1(t-3)','var1(t-2)','var1(t-1)']]
+		predictions = loaded_model.predict([input['weekday'], input['month'], continuas])
+
+		print("Transformando datos")
+		loaded_scaler = load_object('scaler_time_series.pkl')
+		inverted = loaded_scaler.inverse_transform(predictions)
+		inverted = inverted.astype('int32')
+
+		final_predictions = pd.DataFrame(inverted)
+		final_predictions.columns = ['ventas']
+
+		print("Enviar respuesta")
+		responses = jsonify(predictions=final_predictions.to_json(orient="records"))
+		responses.status_code = 200
+		print("Fin de Peticion")
+
+		return (responses)

api_ml/test_api.py

@@ -0,0 +1,42 @@
+import json
+import requests
+import pandas as pd
+import pickle
+from utiles import *
+
+"""Setting the headers to send and accept json responses
+"""
+header = {'Content-Type': 'application/json', \
+                  'Accept': 'application/json'}
+
+# creamos un dataset de pruebas
+df = pd.DataFrame({"unidades": [289,288,260,240,290,255,270,300], 
+					"weekday": [5,0,1,2,3,4,5,0], 
+					"month":   [4,4,4,4,4,4,4,4]})
+
+loaded_scaler = load_object('scaler_time_series.pkl')
+
+reframed = transformar(df, loaded_scaler)
+
+reordenado=reframed[ ['weekday','month','var1(t-7)','var1(t-6)','var1(t-5)','var1(t-4)','var1(t-3)','var1(t-2)','var1(t-1)'] ]
+reordenado.dropna(inplace=True)
+#print(reordenado)
+
+"""Converting Pandas Dataframe to json
+"""
+data = reordenado.to_json(orient='records')
+
+print('JSON para enviar en POST', data)
+
+"""POST <url>/predict
+"""
+resp = requests.post("http://localhost:8000/predict", \
+                    data = json.dumps(data),\
+                    headers= header)
+
+print('status',resp.status_code)
+
+
+print('Respuesta de Servidor')
+print(resp.json())
+