Save and load forecasters¶

Skforecast models can be easily saved and loaded from disk using the pickle or joblib library. Two handy functions, save_forecaster and load_forecaster are available to streamline this process. See below for a simple example.

A forecaster_id has been included when initializing the Forecaster, this may help to identify the target of the model.

✎ Note

Learn how to use forecaster models in production.

Libraries¶

In [1]:

# Libraries
# ==============================================================================
import pandas as pd
from sklearn.ensemble import RandomForestRegressor
from skforecast.ForecasterAutoreg import ForecasterAutoreg
from skforecast.ForecasterAutoregCustom import ForecasterAutoregCustom
from skforecast.utils import save_forecaster
from skforecast.utils import load_forecaster
from skforecast.datasets import fetch_dataset

# Libraries # ============================================================================== import pandas as pd from sklearn.ensemble import RandomForestRegressor from skforecast.ForecasterAutoreg import ForecasterAutoreg from skforecast.ForecasterAutoregCustom import ForecasterAutoregCustom from skforecast.utils import save_forecaster from skforecast.utils import load_forecaster from skforecast.datasets import fetch_dataset

Data¶

In [2]:

# Download data
# ==============================================================================
data = fetch_dataset(
    name="h2o", raw=True, kwargs_read_csv={"names": ["y", "date"], "header": 0}
)
data['date'] = pd.to_datetime(data['date'], format='%Y-%m-%d')
data = data.set_index('date')
data = data.asfreq('MS')

# Download data # ============================================================================== data = fetch_dataset( name="h2o", raw=True, kwargs_read_csv={"names": ["y", "date"], "header": 0} ) data['date'] = pd.to_datetime(data['date'], format='%Y-%m-%d') data = data.set_index('date') data = data.asfreq('MS')

h2o
---
Monthly expenditure ($AUD) on corticosteroid drugs that the Australian health
system had between 1991 and 2008.
Hyndman R (2023). fpp3: Data for Forecasting: Principles and Practice(3rd
Edition). http://pkg.robjhyndman.com/fpp3package/,https://github.com/robjhyndman
/fpp3package, http://OTexts.com/fpp3.
Shape of the dataset: (204, 2)

Save and load forecaster model¶

In [3]:

# Create and train forecaster
# ==============================================================================
forecaster = ForecasterAutoreg(
                 regressor     = RandomForestRegressor(random_state=123),
                 lags          = 5,
                 forecaster_id = "forecaster_001"
             )

forecaster.fit(y=data['y'])
forecaster.predict(steps=3)

# Create and train forecaster # ============================================================================== forecaster = ForecasterAutoreg( regressor = RandomForestRegressor(random_state=123), lags = 5, forecaster_id = "forecaster_001" ) forecaster.fit(y=data['y']) forecaster.predict(steps=3)

Out[3]:

2008-07-01    0.714526
2008-08-01    0.789144
2008-09-01    0.818433
Freq: MS, Name: pred, dtype: float64

In [4]:

# Save model
# ==============================================================================
save_forecaster(forecaster, file_name='forecaster_001.joblib', verbose=False)

# Save model # ============================================================================== save_forecaster(forecaster, file_name='forecaster_001.joblib', verbose=False)

In [5]:

# Load model
# ==============================================================================
forecaster_loaded = load_forecaster('forecaster_001.joblib', verbose=True)

# Load model # ============================================================================== forecaster_loaded = load_forecaster('forecaster_001.joblib', verbose=True)

================= 
ForecasterAutoreg 
================= 
Regressor: RandomForestRegressor(random_state=123) 
Lags: [1 2 3 4 5] 
Transformer for y: None 
Transformer for exog: None 
Window size: 5 
Weight function included: False 
Differentiation order: None 
Exogenous included: False 
Exogenous variables names: None 
Training range: [Timestamp('1991-07-01 00:00:00'), Timestamp('2008-06-01 00:00:00')] 
Training index type: DatetimeIndex 
Training index frequency: MS 
Regressor parameters: {'bootstrap': True, 'ccp_alpha': 0.0, 'criterion': 'squared_error', 'max_depth': None, 'max_features': 1.0, 'max_leaf_nodes': None, 'max_samples': None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 1, 'min_samples_split': 2, 'min_weight_fraction_leaf': 0.0, 'monotonic_cst': None, 'n_estimators': 100, 'n_jobs': None, 'oob_score': False, 'random_state': 123, 'verbose': 0, 'warm_start': False} 
fit_kwargs: {} 
Creation date: 2024-07-29 18:14:22 
Last fit date: 2024-07-29 18:14:22 
Skforecast version: 0.13.0 
Python version: 3.11.5 
Forecaster id: forecaster_001 

In [6]:

# Predict
# ==============================================================================
forecaster_loaded.predict(steps=3)

# Predict # ============================================================================== forecaster_loaded.predict(steps=3)

Out[6]:

2008-07-01    0.714526
2008-08-01    0.789144
2008-09-01    0.818433
Freq: MS, Name: pred, dtype: float64

In [7]:

# Forecaster identifier
# ==============================================================================
forecaster.forecaster_id

# Forecaster identifier # ============================================================================== forecaster.forecaster_id

Out[7]:

'forecaster_001'

Saving and Loading a Forecaster Model with Custom Functions¶

Sometimes external functions are needed when creating a Forecaster object. For example:

• A custom function to create window and custom features.

• A function to reduce the impact of some dates on the model, Weighted Time Series Forecasting.

For your code to work properly, these functions must be available in the environment where the Forecaster is loaded.

In [8]:

# Custom function to create predictors
# ==============================================================================
def create_predictors(y):
    """
    Create first 5 lags of a time series.
    """
    
    lags = y[-1:-6:-1]
    
    return lags

# Custom function to create predictors # ============================================================================== def create_predictors(y): """ Create first 5 lags of a time series. """ lags = y[-1:-6:-1] return lags

In [9]:

# Create and train forecaster
# ==============================================================================
forecaster = ForecasterAutoregCustom(
                 regressor      = RandomForestRegressor(random_state=123),
                 fun_predictors = create_predictors,
                 window_size    = 5
             )

forecaster.fit(y=data['y'])

# Create and train forecaster # ============================================================================== forecaster = ForecasterAutoregCustom( regressor = RandomForestRegressor(random_state=123), fun_predictors = create_predictors, window_size = 5 ) forecaster.fit(y=data['y'])

The save_forecaster function will save all the functions used to create it in different files. In this case, the created module is create_predictors.py.

In [10]:

# Save model and custom function
# ==============================================================================
save_forecaster(
    forecaster, 
    file_name = 'forecaster_custom.joblib', 
    save_custom_functions = True, 
    verbose = False
)

# Save model and custom function # ============================================================================== save_forecaster( forecaster, file_name = 'forecaster_custom.joblib', save_custom_functions = True, verbose = False )

These functions must be imported into the environment where the Forecaster is going to be loaded.

In [11]:

# Load model and custom function
# ==============================================================================
from create_predictors import create_predictors

forecaster_loaded = load_forecaster('forecaster_custom.joblib', verbose=True)

# Load model and custom function # ============================================================================== from create_predictors import create_predictors forecaster_loaded = load_forecaster('forecaster_custom.joblib', verbose=True)

======================= 
ForecasterAutoregCustom 
======================= 
Regressor: RandomForestRegressor(random_state=123) 
Predictors created with function: create_predictors 
Transformer for y: None 
Transformer for exog: None 
Window size: 5 
Weight function included: False 
Differentiation order: None 
Exogenous included: False 
Type of exogenous variable: None 
Exogenous variables names: None 
Training range: [Timestamp('1991-07-01 00:00:00'), Timestamp('2008-06-01 00:00:00')] 
Training index type: DatetimeIndex 
Training index frequency: MS 
Regressor parameters: {'bootstrap': True, 'ccp_alpha': 0.0, 'criterion': 'squared_error', 'max_depth': None, 'max_features': 1.0, 'max_leaf_nodes': None, 'max_samples': None, 'min_impurity_decrease': 0.0, 'min_samples_leaf': 1, 'min_samples_split': 2, 'min_weight_fraction_leaf': 0.0, 'monotonic_cst': None, 'n_estimators': 100, 'n_jobs': None, 'oob_score': False, 'random_state': 123, 'verbose': 0, 'warm_start': False} 
fit_kwargs: {} 
Creation date: 2024-07-29 18:14:23 
Last fit date: 2024-07-29 18:14:23 
Skforecast version: 0.13.0 
Python version: 3.11.5 
Forecaster id: None