Azure 机器学习设计器算法和模块参考Algorithm & module reference for Azure Machine Learning designer
此参考内容提供了 Azure 机器学习设计器中可用的每种机器学习算法和模块的技术背景。This reference content provides the technical background on each of the machine learning algorithms and modules available in Azure Machine Learning designer.
每个模块表示一组可以独立运行并可根据所需输入来执行机器学习任务的代码。Each module represents a set of code that can run independently and perform a machine learning task, given the required inputs. 模块可能包含特定的算法,或者可能执行在机器学习中非常重要的任务,如替换缺少的值或进行统计分析。A module might contain a particular algorithm, or perform a task that is important in machine learning, such as missing value replacement, or statistical analysis.
有关选择算法的帮助,请参阅For help with choosing algorithms, see
提示
在设计器的任何管道中,可以获取有关特定模块的信息。In any pipeline in the designer, you can get information about a specific module. 在模块列表中的模块上方悬停时,或是在模块的右窗格中,选择模块卡上的“了解更多”链接。Select the Learn more link in the module card when hovering on the module in the module list, or in the right pane of the module.
数据准备模块Data preparation modules
功能Functionality | 说明Description | 模块Module |
---|---|---|
数据输入和输出Data Input and Output | 将数据从云源移动到管道中。Move data from cloud sources into your pipeline. 在运行管道时将结果或中间数据写入到 Azure 存储、SQL 数据库或 Hive,或者使用云存储空间在管道之间交换数据。Write your results or intermediate data to Azure Storage, SQL Database, or Hive, while running a pipeline, or use cloud storage to exchange data between pipelines. | 手动输入数据Enter Data Manually 导出数据Export Data 导入数据Import Data |
数据转换Data Transformation | 对数据进行的机器学习独有的操作,例如将数据规范化或装箱、维数缩减以及在各种文件格式间转换数据。Operations on data that are unique to machine learning, such as normalizing or binning data, dimensionality reduction, and converting data among various file formats. | 添加列Add Columns 添加行Add Rows 应用数学运算Apply Math Operation 应用 SQL 转换Apply SQL Transformation 清理缺失数据Clean Missing Data 剪切值Clip Values 转换为 CSVConvert to CSV 转换为数据集Convert to Dataset 转换为指示器值Convert to Indicator Values 编辑元数据Edit Metadata 将数据分组到箱中Group Data into Bins 联接数据Join Data 规范化数据Normalize Data 分区和采样Partition and Sample 删除重复的行Remove Duplicate Rows SMOTESMOTE 选择列转换Select Columns Transform 在数据集中选择列Select Columns in Dataset 拆分数据Split Data |
特征选择Feature Selection | 选择用于构建分析模型的相关有用特征的子集。Select a subset of relevant, useful features to use in building an analytical model. | 基于筛选器的特征选择Filter Based Feature Selection 排列特征重要性Permutation Feature Importance |
统计函数Statistical Functions | 提供与数据科学相关的各种统计方法。Provide a wide variety of statistical methods related to data science. | 汇总数据Summarize Data |
机器学习算法Machine learning algorithms
功能Functionality | 说明Description | 模块Module |
---|---|---|
回归Regression | 预测值。Predict a value. | 提升决策树回归Boosted Decision Tree Regression 决策林回归Decision Forest Regression 快速林分位回归Fast Forest Quantile Regression 线性回归Linear Regression 神经网络回归Neural Network Regression 泊松回归Poisson Regression |
群集Clustering | 将数据分到一组。Group data together. | K 均值聚类分析K-Means Clustering |
分类Classification | 预测类。Predict a class. 从二进制(双类)或多类算法中进行选择。Choose from binary (two-class) or multiclass algorithms. | 多类提升决策树Multiclass Boosted Decision Tree 多类决策林Multiclass Decision Forest 多类逻辑回归Multiclass Logistic Regression 多类神经网络Multiclass Neural Network “一对多”多类One vs. All Multiclass “一对一个多类One vs. One Multiclass 双类平均感知器Two-Class Averaged Perceptron 双类提升决策树Two-Class Boosted Decision Tree 双类决策林Two-Class Decision Forest 双类逻辑回归Two-Class Logistic Regression 双类神经网络Two-Class Neural Network 双类支持向量机Two Class Support Vector Machine |
用于构建和评估模型的模块Modules for building and evaluating models
Web 服务Web service
了解 Azure 机器学习设计器中的实时推理所需的 Web 服务模块。Learn about the web service modules which are necessary for real-time inference in Azure Machine Learning designer.
错误消息Error messages
了解在 Azure 机器学习设计器中使用模块时可能会遇到的错误消息和异常代码。Learn about the error messages and exception codes you might encounter using modules in Azure Machine Learning designer.