sql-machine-learning/sqlflow
Sqlflow
sqlflow 是一个 SQL 机器学习引擎与编排器，旨在通过扩展的 SQL 查询语法来训练、部署与解释机器学习模型。它通过将数据库引擎连接到外部机器学习工具包，实现了数据库内机器学习，允许用户直接通过查询定义训练数据集与超参数。
该系统作为预测接口与可解释性工具运行。它允许通过在标准 SQL 语句中调用模型函数，对数据库记录生成分类与预测，并提供了一个工作流来解释特定特征如何影响模型决策。
Features
- SQL-Based Machine Learning - Defines machine learning training and inference parameters using a custom SQL query syntax.
- In-Database Machine Learning - Enables the training and execution of machine learning models directly within the database engine using SQL.
- Machine Learning Toolkits - Integrates with specialized machine learning toolkits to facilitate model training and prediction execution.
- Machine Learning Training - Supports building machine learning models by defining datasets and hyperparameters via SQL syntax.
- Model Inference - Enables model inference by calling prediction functions directly within SQL select statements.
- Model Predictions - Generates predicted values and classifications from trained models using standard SQL queries.
- Model Explainability - Uses query extensions to quantify the influence of specific features on model predictions.
- AI and Machine Learning - SQL-based machine learning engine.
- Machine Learning Operations - Brings machine learning capabilities to SQL, enabling model training and prediction using SQL syntax.
- Data Processing Tools - Bridge between SQL queries and machine learning workflows.
