Enhance Time Series Forecasting with NeuralProphet and ZenML

Integrate NeuralProphet's powerful time series forecasting capabilities into your ZenML pipelines for seamless and efficient modeling of temporal data. This integration enables you to leverage NeuralProphet's state-of-the-art algorithms and intuitive API within the structured framework of ZenML, streamlining your time series modeling workflow.

Features with ZenML

Seamless Integration: Effortlessly incorporate NeuralProphet models into your ZenML pipelines, ensuring a smooth and unified workflow.
Reproducibility: Leverage ZenML's versioning and tracking capabilities to ensure reproducibility and traceability of your NeuralProphet models.

‍

Main Features

Flexible time series modeling with support for trends, seasonality, and holidays
Intuitive API for defining and training models
Automatic hyperparameter tuning for optimized model performance
Built-in model evaluation and visualization tools
Ability to incorporate external regressors and custom loss functions

‍

How to use ZenML with

NeuralProphet


from zenml import pipeline, step
import pandas as pd
from neuralprophet import NeuralProphet

@step
def load_data() -> pd.DataFrame:
    # Load your time series data here
    data = pd.read_csv('time_series_data.csv')
    return data

@step
def trainer_step(data: pd.DataFrame) -> NeuralProphet:
    model = NeuralProphet()
    model.fit(data, freq="D")
    return model

@pipeline
def time_series_pipeline():
    data = load_data()
    trainer_step(data)


if __name__ == "__main__":
    time_series_pipeline()

This code example demonstrates a simple ZenML pipeline that integrates NeuralProphet for time series forecasting. The pipeline consists of two steps: