tessl/pypi-prophet

Automatic forecasting procedure for time series data with strong seasonal effects and multiple seasons of historical data

—

Pending

Overview

Eval results

Files

Serialization

Name: tessl/pypi-prophet
Author: tessl

Model persistence functionality for saving and loading trained Prophet models using JSON serialization with support for all model components including parameters, history, and Stan backend state.

Capabilities

JSON Serialization

Save and load Prophet models using JSON format.

def model_to_json(model):
    """
    Serialize a Prophet model to JSON string.

    Parameters:
    - model: Prophet, fitted Prophet model

    Returns:
    - str, JSON string representation of the model
    """

def model_from_json(model_json):
    """
    Deserialize a Prophet model from JSON string.

    Parameters:
    - model_json: str, JSON string representation of a Prophet model

    Returns:
    - Prophet, reconstructed Prophet model
    """

Dictionary Serialization

Convert Prophet models to/from Python dictionaries.

def model_to_dict(model):
    """
    Convert a Prophet model to a dictionary.

    Parameters:
    - model: Prophet, fitted Prophet model

    Returns:
    - dict, dictionary representation of the model
    """

def model_from_dict(model_dict):
    """
    Recreate a Prophet model from a dictionary.

    Parameters:
    - model_dict: dict, dictionary representation of a Prophet model

    Returns:
    - Prophet, reconstructed Prophet model
    """

Serialization Constants

Internal constants defining which model attributes are serialized.

SIMPLE_ATTRIBUTES = [
    'growth', 'n_changepoints', 'changepoint_range', 'yearly_seasonality',
    'weekly_seasonality', 'daily_seasonality', 'seasonality_mode',
    'seasonality_prior_scale', 'holidays_prior_scale', 'changepoint_prior_scale',
    'mcmc_samples', 'interval_width', 'uncertainty_samples', 'scaling',
    'holidays_mode'
]
# List of simple model attributes to serialize

PD_SERIES = ['y_scale', 'start', 'y_min', 'y_max', 't_scale']
# Pandas Series attributes

PD_TIMESTAMP = ['start']
# Pandas Timestamp attributes  

PD_TIMEDELTA = ['t_scale']
# Pandas Timedelta attributes

PD_DATAFRAME = [
    'history', 'holidays', 'train_holiday_names', 'train_component_cols',
    'component_modes', 'seasonalities', 'extra_regressors'
]
# Pandas DataFrame attributes

NP_ARRAY = ['changepoints_t', 'train_component_cols']
# NumPy array attributes

ORDEREDDICT = ['seasonalities', 'extra_regressors']
# OrderedDict attributes

Usage Examples

Save and Load Model

from prophet import Prophet
from prophet.serialize import model_to_json, model_from_json

# Train a model
model = Prophet()
model.fit(df)

# Save model to JSON string
model_json = model_to_json(model)

# Save to file
with open('prophet_model.json', 'w') as f:
    f.write(model_json)

# Load from file
with open('prophet_model.json', 'r') as f:
    model_json = f.read()

# Recreate model from JSON
loaded_model = model_from_json(model_json)

# Use loaded model for prediction
forecast = loaded_model.predict(future)

Dictionary-based Serialization

from prophet.serialize import model_to_dict, model_from_dict
import pickle

# Convert model to dictionary
model_dict = model_to_dict(model)

# Save dictionary using pickle
with open('prophet_model.pkl', 'wb') as f:
    pickle.dump(model_dict, f)

# Load and reconstruct
with open('prophet_model.pkl', 'rb') as f:
    loaded_dict = pickle.load(f)

reconstructed_model = model_from_dict(loaded_dict)

Model Validation After Loading

import numpy as np

# Verify model integrity after loading
original_forecast = model.predict(future)
loaded_forecast = loaded_model.predict(future)

# Check if forecasts are identical
forecast_match = np.allclose(
    original_forecast['yhat'].values,
    loaded_forecast['yhat'].values
)
print(f"Forecasts match: {forecast_match}")

Install with Tessl CLI