azure机器学习

介绍 (Introduction)

Let us see how Azure ML studio can be used to create machine learning models and how to consume them in this series. As we discussed during the data mining series, we identified the challenges in the predictions in data. In the Azure Machine learning platform, machine learning workflows can be defined in easy scale models in the cloud environment. Today will be looking at how datasets can be uploaded.

让我们看看如何使用Azure ML studio创建机器学习模型，以及如何在本系列文章中使用它们。 正如我们在数据挖掘系列中讨论的那样，我们确定了数据预测中的挑战。 在Azure机器学习平台中，可以在云环境中的轻松规模模型中定义机器学习工作流。 今天将研究如何上传数据集。

MLOPS (MLOPS)

DevOps is a very familiar tool among IT practitioners nowadays so the development and operation teams can get-together and work to the success of the project. Similarly, in Machine Learning, there are different teams such as data scientists, data engineers, and development teams working on machine learning projects to work on models and consuming them. Azure machine learning can utilize the MLOps model to build high quality and scalable machine learning models that are equivalent to the DevOps.

DevOps是当今IT从业人员非常熟悉的工具，因此开发和运营团队可以齐心协力，为项目的成功而努力。 同样，在机器学习中，有不同的团队，例如数据科学家，数据工程师和开发团队，负责机器学习项目，以处理模型并使用它们。 Azure机器学习可以利用MLOps模型来构建与DevOps等效的高质量和可扩展机器学习模型。

Further, if you look at the machine learning development life cycle, you need multiple tasks such as,

此外，如果您查看机器学习开发生命周期，则需要执行多个任务，例如，

• Pre-processing data 预处理数据
• Preparing data 准备资料
• Developing candidate ML models 开发候选ML模型
• Evaluating candidate ML models 评估候选ML模型
• Choosing an ML model 选择ML模型
• Deploying the selected Machine learning model 部署选定的机器学习模型
• Consuming the ML model 消耗机器学习模型

Azure Machine Learning will provide different users at different tasks in the development life cycle of machine learning.

Azure机器学习将在机器学习的开发生命周期中为不同的用户提供不同的任务。

Azure机器学习 (Azure Machine Learning )

To facilitate all the above tasks, you can use the Azure Machine Learning Studio which is the browser-based workbench for Machine Learning. You can create your free account to try out many features of the Azure Machine Learning Platform. For example, in the free account maximum storage is 10 GB whereas there is no limit in the paid account. Apart from the storage limitation, the free account will execute on a single node and the paid account is running on multiple nodes. Apart from those limitations, the free account does not require an Azure subscription.

为了完成上述所有任务，您可以使用Azure Machine Learning Studio，这是用于机器学习的基于浏览器的工作台。 您可以创建一个免费帐户来试用Azure机器学习平台的许多功能。 例如，在免费帐户中，最大存储量为10 GB，而在付费帐户中则没有限制。 除存储限制外，免费帐户将在单个节点上执行，而付费帐户将在多个节点上运行。 除了这些限制，免费帐户不需要Azure订阅。

You can look at the details of limitations from the following URL as these limits will change from time to time:

您可以从以下网址查看限制的详细信息，因为这些限制会不时更改：

https://azure.microsoft.com/en-us/pricing/details/machine-learning-studio/

https://azure.microsoft.com/zh-CN/pricing/details/machine-learning-studio/

After an account is created, you can log in to https://studio.azureml.net/ where it provides a lot of videos and important documentation for a novice user.

创建帐户后，您可以登录https://studio.azureml.net/ ，该帐户为新手提供了许多视频和重要文档。

If you are creating a machine learning resource form the Azure Portal, in the AI + Machine Learning category choose the machine learning resource as shown in the below image:

如果要通过Azure门户创建机器学习资源，请在“ AI +机器学习”类别中选择机器学习资源，如下图所示：

Let us provide the basic details for the machine learning recourse as shown in the below image:

让我们提供机器学习资源的基本细节，如下图所示：

In the above screen, you need to provide the Azure subscription and resource group. The region will select a data center that will execute your machine learning projects. Unlike most of the other azure service, Azure machine learning does not exist in all the data centers around the world.

在上面的屏幕中，您需要提供Azure订阅和资源组。 该地区将选择一个数据中心来执行您的机器学习项目。 与大多数其他Azure服务不同，Azure机器学习并不存在于全球所有数据中心中。

Next is to create tags for the azure machine learning services as shown in the below screenshot:

接下来是为azure机器学习服务创建标签，如以下屏幕截图所示：

The above tags will be used for billing and costing purposes.

以上标签将用于计费和成本核算。

Let us use the https://studio.azureml.net for the demo purposes.

让我们使用https://studio.azureml.net进行演示。

Azure ML Studio (Azure ML Studio)

Azure Machine Learning Studio has multiple options as shown in the below figure:

Azure Machine Learning Studio具有多个选项，如下图所示：

A Project is a collection of multiple assets such as Datasets, experiments, etc and a project need a project name as description as shown in the below screenshot:

一个项目是多个资产的集合，例如数据集，实验等，并且一个项目需要一个项目名称作为说明，如以下屏幕快照所示：

Previously created data sets and experiments etc can be added to the project.

先前创建的数据集和实验等可以添加到项目中。

Experiments are the models that will be created by data scientists to predict and model the data. When a new Experiment is created, you can choose from the existing sample experiment templates or if you want to start from the beginning, you can use the Blank Experiment:

实验是数据科学家将创建的用于预测和建模数据的模型。 创建新实验后 ，您可以从现有的示例实验模板中进行选择，或者如果要从头开始，则可以使用空白实验 ：

When the Blank Experiment is selected, the following screen can be seen in the Azure ML studio:

选择“ 空白实验”后 ，可以在Azure ML Studio中看到以下屏幕：

You can drag and drop the experiment items that will be discussed in detail during the upcoming article in this series.

您可以拖放将在本系列下一篇文章中详细讨论的实验项目。

Web Services can be created so that it can be consumed by the different users in different applications. These web services can be consumed from Microsoft Excel as well that will be looked at during this article series.

可以创建Web服务，以便不同应用程序中的不同用户可以使用它。 这些Web服务也可以从Microsoft Excel中使用，这将在本系列文章中进行介绍。

Data Sets are the data that you will be working on. For you to try out options in Azure Machine Learing studio, there are a lot of real-world data samples as shown in the below screen:

数据集是您将要处理的数据。 为了让您在Azure Machine Learing Studio中试用选项，有很多真实的数据示例，如以下屏幕所示：

We will be using some of these data samples in future articles to demonstrate different machine learning techniques. You can download the data set or you can add a dataset to a project.

在以后的文章中，我们将使用其中一些数据样本来演示不同的机器学习技术。 您可以下载数据集，也可以将数据集添加到项目中。

If you have a data set, you can upload the data set using New option in the Dataset option as shown in the below screenshot:

如果有数据集，则可以使用“数据集”选项中的“ 新建”选项上载数据集 ，如以下屏幕截图所示：

For the above dataset import, iris sample data set of WEKA is used. Apart from arff files, CSV, TSV, text files, Zip files can be uploaded to the dataset.

对于上面的数据集导入，使用WEKA的虹膜样本数据集。 除了arff文件，CSV，TSV，文本文件，Zip文件，还可以上传到数据集。

You can now observe your data sets in the left-hand side list as shown in the below screen:

现在，您可以在左侧列表中观察数据集，如以下屏幕所示：

You can find the basic statistical properties of the data set by drag and drop the data set to an experiment:

您可以通过将数据集拖放到实验中来找到数据集的基本统计属性：

You can view the statistics of a dataset by selecting the Visualize option as shown in the below screen:

您可以通过选择“ 可视化”选项来查看数据集的统计信息，如以下屏幕所示：

After visualize option is selected, the statistical properties of data can be viewed.

选择可视化选项后，可以查看数据的统计属性。

By clicking any column, you can view the statistical details such as mean, median, minimum, maximum, standard deviation, Unique values, missing values as shown below:

通过单击任何列，您可以查看统计信息，例如平均值，中位数，最小值，最大值，标准偏差，唯一值，缺失值，如下所示：

You can either view the data in bar charts or boxplots by choosing the necessary option as shown in the below screenshot:

您可以通过选择必要的选项来查看条形图或箱图中的数据，如以下屏幕截图所示：

In the above diagram, data is distributed to 10 bins that is the default value. The bin value can be modified according to your need. The frequency of values can be converted to the log scale as well.

在上图中，数据被分配到默认值的10个bin中。 bin值可以根据需要进行修改。 值的频率也可以转换为对数刻度。

Another important feature of the Azure ML Studio is the ability to compare values so that you can get an idea of your data set. For example, let us say we want to find the distribution of the petal width for different classes of iris flowers, you can choose the class as the comparable attribute and the following screen will be visible:

Azure ML Studio的另一个重要功能是可以比较值，以便您可以了解数据集。 例如，假设我们要查找不同类别的鸢尾花的花瓣宽度分布，您可以选择该类别作为可比较属性，然后将显示以下屏幕：

By looking at the above screen, data distribution can be easily identified.

通过查看以上屏幕，可以轻松识别数据分布。

In the Trained model tab, you will find the models that were trained and in the Setting tab, you will have the options to manage your projects:

在“ 训练的模型”选项卡中，您将找到受过训练的模型，在“ 设置”选项卡中，您将具有管理项目的选项：

You can view the details for the workspace from the Name tab whereas the Users tab provides you the option of sharing the experiments and projects with your peers to support the MLOps as we discussed previously. Those users should have windows live account and when they are added, they will be notified via an email.

您可以从“ 名称”选项卡中查看工作区的详细信息，而“ 用户”选项卡为您提供了与对等方共享实验和项目以支持MLOps的选项，如上所述。 这些用户应具有Windows Live帐户，并且在添加时将通过电子邮件通知他们。

Data Gateways provide you with the option of accessing the on-premises database. However, your on-premises database feature is not available for the “Free” tier and you need to upgrade to the “Standard” tier to access data from on-premises.

数据网关为您提供了访问本地数据库的选项。 但是，本地数据库功能不适用于“免费”层，您需​​要升级到“标准”层才能从本地访问数据。

结论 (Conclusion)

Azure ML Studio provides MLOps capabilities for the different users in machine learning projects. Apart from building scalable ML models, users can upload their own data set and observe the statistical properties of the dataset.

Azure ML Studio为机器学习项目中的不同用户提供了MLOps功能。 除了构建可扩展的ML模型外，用户还可以上传自己的数据集并观察数据集的统计属性。

目录 (Table of contents)

Introduction to Azure Machine Learning using Azure ML Studio

Data Cleansing in Azure Machine Learning

Prediction in Azure Machine Learning

使用Azure ML Studio的Azure机器学习简介

Azure机器学习中的数据清理

Azure机器学习中的预测

翻译自: https://www.sqlshack.com/introduction-to-azure-machine-learning-using-azure-ml-studio/

azure机器学习