def predict_delay(airport_code):
firebase_source = mapper.get_source_firebase()
airport_status = firebase_source.get_airport(airport_code)
cleaned_data = utils.get_clean_data(airport_status)
res = predictor.predict(airport_code)
cleaned_data["prediction"] = bool(res[0])
return jsonify(cleaned_data)
def predict_delay(airport_code):
firebase_source = mapper.get_source_firebase()
airport_status = firebase_source.get_airport(airport_code)
cleaned_data = utils.get_clean_data(airport_status)
res = predictor.predict(airport_code)
cleaned_data["prediction"] = bool(res[0])
return jsonify(cleaned_data)
def get_airport_statuses():
firebase_source = mapper.get_source_firebase()
airports = firebase_source.get_all()
results = []
for airport_code, status in airports.items():
try:
results.append(utils.get_clean_data(status))
except:
pass
results = {"items": results}
return jsonify(results)
def get_airport_statuses():
firebase_source = mapper.get_source_firebase()
airports = firebase_source.get_all()
results = []
for airport_code, status in airports.items():
try:
results.append(utils.get_clean_data(status))
except:
pass
results = {"items":results}
return jsonify(results)
def clean_data(self):
uncleaned = self.dump_firebase.get_all()
uncleaned = uncleaned or {}
firebase_clean = FirebaseApplication(app.config["FIREBASE_CLEAN"], None)
cleaned = firebase_clean.get("/", None)
for airport_code, date_status_dict in uncleaned.items():
if type(date_status_dict) != dict:
print airport_code, date_status_dict
continue
for date, event in date_status_dict.items():
if airport_code in cleaned and date in cleaned[airport_code]:
continue
filtered_data = utils.get_clean_data(event)
try:
firebase_clean.put(url="/"+airport_code, name=date, data=filtered_data)
print "Cleaned", airport_code, date
except Exception as e:
print "Could not clean", airport_code, date, event
print e
def get_flight_info_from_firebase(self):
response = self.source_firebase.get("", None)
updated = parser.parse(response["_updated"])
if updated <= self.last_updated + self.delta:
raise Exception("Not time to sync yet.")
print "Syncing Airpot Delays from", updated.isoformat()
self.last_updated = updated
update_str = updated.isoformat()
firebase_clean = FirebaseApplication(app.config["FIREBASE_CLEAN"], None)
for airport_code, info in response.items():
# this is not an airport :)
if airport_code == "_updated": continue
try:
self.dump_firebase.firebase.put(url="/"+airport_code, name=update_str, data=info)
filtered_data = utils.get_clean_data(info)
firebase_clean.put(url="/"+airport_code, name=update_str, data=filtered_data)
except Exception as e:
print "Something failed", e
self.dump_firebase.firebase.put("/", "last_updated", update_str)
)
# The Adam optimizer is used.
optimizer = keras.optimizers.Adam(learning_rate=lr_schedule)
# The model is trained using accuracy, and the loss function is binary crossentropy.
model.compile(optimizer=optimizer,
loss="binary_crossentropy",
metrics=["accuracy"])
return model
if __name__ == "__main__":
features = ["Pseudo-Log-Return", "Trend Indicator"]
label = ["Direction"]
# Get each set of data.
df_train, df_valid, df_test = get_clean_data()
Xtrain = df_train[features].values
Ytrain = df_train[label].values.ravel()
Xvalid = df_valid[features].values
Yvalid = df_valid[label].values.ravel()
Xtest = df_test[features].values
Ytest = df_test[label].values.ravel()
# Store the data for each set in the corresponding feature and output vectors.
# Preprocess the data. https://scikit-learn.org/stable/modules/preprocessing.html
scale = MinMaxScaler(feature_range=(0, 1))
Xtrain = scale.fit_transform(Xtrain)
Ytrain = scale.fit_transform(Ytrain.reshape(-1, 1))
Xvalid = scale.fit_transform(Xvalid)
Yvalid = scale.fit_transform(Yvalid.reshape(-1, 1))
Xtest = scale.fit_transform(Xtest)
