def test_regression_metrics():
logger.info("Testeando métricas de evaluación para regresión...")
# Ejemplos para testear regresión
dim = 100 # Tamaño del muestreo
x = np.linspace(-4, 4, dim) # Eje x
clean = np.sinc(x) # Función sinc (labels)
np.random.seed(123)
noise = np.random.uniform(0, 0.1, dim) # Ruido para ensuciar la sinc (error en la predicción)
signal = clean + noise # Señal resultante del ruido aditivo (predict)
metrics = RegressionMetrics(zip(signal, clean))
assert np.allclose(metrics.r2(), 0.9708194315829859, rtol=1e-4)
assert np.allclose(metrics.explained_variance(), 0.9943620888461356, rtol=1e-4)
assert np.allclose(metrics.mse(), 0.0031164269743473839, rtol=1e-4)
assert np.allclose(metrics.rmse(), 0.0558249673027, rtol=1e-4)
assert np.allclose(metrics.mae(), 0.0501428880, rtol=1e-4)
assert np.allclose(metrics.rmae(), 0.2239260770, rtol=1e-4)
logger.info("OK")
hits_valid = neural_net.fit(train,
valid,
valid_iters=1,
mini_batch=20,
parallelism=0,
stops=global_stops,
optimizer_params=optimizer_params,
measure='R2',
keep_best=True,
reproducible=False)
hits_test, predict = neural_net.evaluate(test, predictions=True)
logger.info("Hits en test: %12.11f", hits_test)
# --4) Evaluacion de desempeño
logger.info("Metricas de evaluación en clasificacion: ")
labels = map(lambda lp: float(lp.label), test.collect())
metrics = RegressionMetrics(zip(predict, labels))
print "MSE: ", metrics.mse()
print "RMSE: ", metrics.rmse()
print "MAE: ", metrics.mae()
print "RMAE: ", metrics.rmae()
print "R-cuadrado: ", metrics.r2()
print "Explained Variance: ", metrics.explained_variance()
print zip(predict, labels)[:10]
# --5) Ploteos
plot_fitting(neural_net)
logger.info("Configuracion usada: %s", os.linesep+str(net_params))
# -- 3) Construcción y ajuste de red neuronal
neural_net = NeuralNetwork(net_params)
logger.info("Entrenando red neuronal ...")
hits_valid = neural_net.fit(train, valid, valid_iters=1, mini_batch=20, parallelism=0,
stops=global_stops, optimizer_params=optimizer_params, measure='R2',
keep_best=True, reproducible=False)
hits_test, predict = neural_net.evaluate(test, predictions=True)
logger.info("Hits en test: %12.11f", hits_test)
# --4) Evaluacion de desempeño
logger.info("Metricas de evaluación en clasificacion: ")
labels = map(lambda lp: float(lp.label), test.collect())
metrics = RegressionMetrics(zip(predict, labels))
print "MSE: ", metrics.mse()
print "RMSE: ", metrics.rmse()
print "MAE: ", metrics.mae()
print "RMAE: ", metrics.rmae()
print "R-cuadrado: ", metrics.r2()
print "Explained Variance: ", metrics.explained_variance()
print zip(predict, labels)[:10]
# --5) Ploteos
plot_fitting(neural_net)