autoforecast 0.0.9

AutoML time series forecasting

Auto Forecast

AutoML library for time series forecasting

Quick start

Upgrade pip

$ pip install pip --upgrade

Install autoforecast

$ pip install autoforecast

Try it out with your own dataset

1. Preprocessing

from Autoforecast.preprocessing import preprocessing

X_train, y_train, X_test, y_test = preprocessing(
    df=df,
    target_name=['sales'],
    categoricals=['store_id', 'dpt_id', 'holiday', 'zipcode', ...],
    numericals=['employee_num', 'store_surface', ...],
    date_col=['date'],
    train_size=0.8,
    engineering=True,
    selection=True
)

2. Fitting and predicting

from autoforecast.automl import AutoForecast


model = AutoForecast()

print('Autoforecast() model fitting...')
model.fit(X_train=X_train, y_train=y_train)

print('Autoforecast() model predicting...')

y_pred = model.predict(X_test=X_test)
print(f'y_pred={y_pred})

Run the example function

from autoforecast.examples import autoforecast_bitcoin


autoforecast_bitcoin.run()

Fetch historical cryptocurrency data

This function is a wrapper of https://developers.coinbase.com/api/v2#prices

• n: integer, number of days we want since today
• type: str, ['buy', 'sell', 'spot']
• currency_pair: str, crypto & currency

from autoforecast.datasets.import_bitcoin_price import get_price_for_last_n_days


crypto_df = get_price_for_last_n_days(
    n=1, type='spot', currency_pair='BTC-USD')
)