课程

Designing Forecasting Pipelines for Production

高级技能水平

更新时间 2025年12月

Learn how to design, automate, and monitor scalable forecasting pipelines in Python.

免费开始课程

PythonMachine Learning

4小时

16 视频

53 道练习

4,000 XP

成就证明

深受数千家公司学习者的喜爱

需要团队培训？

企业版试用

课程描述

Learn how to design, automate, and monitor scalable forecasting pipelines in Python. This advanced course walks you through the entire production workflow - from sourcing data and training models to deployment and monitoring - using tools like MLflow and Airflow.You'll start by connecting to live data sources and building your first forecast with U.S. electricity demand data. Next, you'll discover experimentation fundamentals, including backtesting, evaluation, and model registration using MLflow.Then you'll build automated forecasting pipelines with ETL processes, model registration, and Airflow orchestration. Finally, you'll learn production deployment essentials, including monitoring pipeline health, detecting model drift, and maintaining forecasting systems in real-world environments.

先决条件

Introduction to Apache Airflow in Python Introduction to MLflow Time Series Analysis in Python

1

General Architecture

Learn how to connect to live data sources and prepare time series data for forecasting. You’ll pull hourly electricity demand data from the U.S. EIA API and build your first forecast.

Productionizing your forecast model

Forecasting project architecture

Benefits of experimentation frameworks

Reviewing the input data

Making an API request

Preparing and visualizing the data

Interpreting time series visualizations

Working with a forecast object

Forecasting with ML Models

Evaluating forecast performance

Visualizing forecast performance

2

Experimentation

Discover the fundamentals of experimentation, including backtesting, evaluation, and model registration using MLflow!

Experimentation

Forecast evaluation & experimentation

Exploring experimentation process

Workflow of experimentation

Backtesting

Defining the forecasting models

Training models with backtesting

Visualizing forecast results

Model Scoring

Transforming backtesting output

Building a model leaderboard

Model Registration with MLflow

Logging tuned models

Searching experiment results

3

Setting Automation

Learn how to build automated forecasting pipelines that refresh data and predictions daily. You'll set up ETL processes, register models with MLflow, and orchestrate everything with Airflow. Create a production-ready system with data validation and logging to monitor pipeline health.

Pipeline architecture

Defining the forecasting pipeline

Fitting the model

Registering the model

Data ingestion

Data parity checks

Defining the schema

Performing data validation

Setting up the DAG

Creating the DAG

Defining pipeline execution order

Logs

Monitoring ETL logs

Monitoring forecasting logs

4

From Deployment to Production

Discover the essentials of production deployment, from monitoring the pipeline health to detecting model drift. You'll learn best practices for reproducibility, scaling, and maintaining forecasting systems in real-world environments.

Pipeline Deployment to Production

What could go wrong?

Monitoring vs Observability

Pipeline Monitoring

Categorizing API risks

Building a mitigation plan

Model Drift

Calculating a threshold and moving averages

Identifying model drift

Best practices

Development best practices

Scaling forecasting pipelines

Congratulations!

Designing Forecasting Pipelines for Production

课程完成

获得成就证明

将此证书添加到您的 LinkedIn 档案、简历或履历中
在社交媒体和绩效评估中分享立即注册

加入超过19百万学习者，今天就开始Designing Forecasting Pipelines for Production！

通过 DataCamp for Mobile 提升您的数据技能

随时随地通过我们的移动课程和每日 5 分钟编程挑战提升技能。