def forecast_settings(message):
# Initial forecast settings - the first time the user sends forecast settings through the app - will use this value in forecastr method
build_settings = 'initial'
# store message['data'] into a df called data
data = message['data']
# Keep Original Data in Exisiting Structure
original_dataset = data[1]['data'][1]['data']
#print("******************** ORIGINAL DATASET *****************************")
#print(original_dataset)
#print("******************** ORIGINAL DATASET *****************************")
# Extract info from forecast_settings message
time_series_data = pd.DataFrame(data[1]['data'][1]['data'])
forecast_settings = data[0]
freq = data[2]
column_headers = data[1]['data'][0]
# Format the date and metric unit
time_unit = column_headers[0]
print(time_unit)
time_series_data[time_unit] = time_series_data[time_unit].apply(
lambda x: pd.to_datetime(str(x)))
metric = column_headers[1]
# y (aka as "the original data for the metric being forecasted") will be used in the chartjs line graph
y = time_series_data[metric].tolist()
# Use Facebook Prophet through forecastr method
forecast = forecastr(time_series_data, forecast_settings, column_headers,
freq, build_settings)
# Need to convert forecast back into a list / array for y, y_hat and date so it can be properly graphed with chartjs
y_hat = forecast[0]
dates = forecast[1]
model = forecast[2]
csv_export = forecast[3]
forecasted_vals = forecast[4]
forecasted_vals_mean = forecast[5]
# Send data back to the client
data_back_to_client = [
dates, y_hat, y, forecast_settings, column_headers, freq,
original_dataset, csv_export, forecasted_vals, forecasted_vals_mean
]
#print(data_back_to_client)
emit('render_forecast_chart', {'data': data_back_to_client})
def update_chart(message):
# This is an update to the initial forecast settings. The user has changed their settings on Step 3, so we set build_settings to update.
build_settings = 'update'
data = message['data']
### Setup variables for use in the forecastr method
time_series_data = data[4]
original_dataset = time_series_data
time_series_data = pd.DataFrame(time_series_data)
#print("********* TIME SERIES DF ****************")
#print(time_series_data.head())
#print("********* TIME SERIES DF ****************")
forecast_settings = data[1]
column_headers = data[2]
freq = data[3]
# Dimension and Metric
time_unit = column_headers[0]
metric = column_headers[1]
# Make sure time_unit is converted to datetime in order to join in helper_v3
time_series_data[time_unit] = time_series_data[time_unit].apply(
lambda x: pd.to_datetime(str(x)))
#print([time_unit,metric])
# Original Data
y = time_series_data[metric].tolist()
# Use Facebook Prophet through forecastr method
forecast = forecastr(time_series_data, forecast_settings, column_headers,
freq, build_settings)
# Need to convert forecast back into a list / array for y, y_hat and date so it can be properly graphed with chartjs
y_hat = forecast[0]
dates = forecast[1]
model = forecast[2]
csv_export = forecast[3]
forecasted_vals = forecast[4]
forecasted_vals_mean = forecast[5]
# Send data back to the client - took out original dataset
data_back_to_client = [
dates, y_hat, y, forecast_settings, column_headers, freq,
original_dataset, csv_export, forecasted_vals, forecasted_vals_mean
]
emit('render_forecast_chart', {'data': data_back_to_client})