Fixing times based on UTC

morphemic-forecasting-eshybrid/forecasting/eshybrid.py

@@ -88,26 +88,28 @@ class ESHybrid(morphemic.handler.ModelHandler,messaging.listener.MorphemicListen
     def on_schedule(self, times):
 
         for m in self.metrics:
-
             predictions = self.model.predict(
                 self.application,
                 m,
                 times
             )
 
-            for t in times:
-                logging.debug("Sending prediction for time %s(%s) " % (datetime.datetime.fromtimestamp(t), t))
+            for index,row in predictions.iterrows():
+                t = row['ds']
+                payload = {
+                    "metricValue": row['y_hat'],
+                    "timestamp": int(t.strftime('%s')),
+                    "probability": 0.98,
+                    "confidence_interval": [float(8),float(15)],
+                    "predictionTime": int(time.time()),
+                }
+                logging.debug("Sending prediction for time %s => %s " % (t, payload) )
                 self.connector.send_to_topic(
                     "intermediate_prediction.eshybrid.%s" % m,
-                    {
-                        "metricValue": predictions[t],
-                        "timestamp": int(time.time()),
-                        "probability": 0.98,
-                        "confidence_interval": [float(8),float(15)],
-                        "predictionTime":t,
-                    }
+                    payload
                 )
 
+
     def _train_model(self):
 
         self.dataset.make()

morphemic-forecasting-eshybrid/morphemic/model.py

 import logging
 import threading
 import uuid
-
+from datetime import datetime
 import pandas as pd
 
 from ESRNN import ESRNN
@@ -63,11 +63,14 @@ class Model:
     def __init__(self, handler=False) -> None:
         self._handler= handler
         # integrated here
-        self._model = False
+        self._model = None
 
     def _new_model(self) -> UUIDModel:
         _logger.debug("Training new model")
         config = configuration.get_config('Hourly')
+        config.get('train_parameters').update({
+            'batch_size':1
+        })
         model = UUIDModel(uuid.uuid4(),config)
         return model
 
@@ -95,7 +98,7 @@ class Model:
         m_pd_test_x['x'] = m
 
 
-        return beautify(m_pd_train_x),beautify(m_pd_train_y), beautify(m_pd_test_x) , None #beautify(m_pd_test_y)
+        return beautify(m_pd_train_x),beautify(m_pd_train_y), beautify(m_pd_test_x) , None
 
     def _retrain(self, metrics, data):
 
@@ -133,10 +136,13 @@ class Model:
             _logger.error("No model trained yet")
             return
 
-        m_pd = pd.DataFrame(data=times, columns=['ds'])
+        m_pd = pd.DataFrame(data=[ datetime.fromtimestamp(x) for x in times ], columns=['ds'])
         m_pd.insert(0,'unique_id',application)
         m_pd.insert(1,'x',metric)
 
+        m_pd['ds'] = pd.to_datetime(m_pd['ds'], utc = True)
+
+
         ret= self._model.model_for_metric(metric).predict(m_pd)
 
         return ret
\ No newline at end of file