当前工作簿宏 被自动删除_会自动删除我的工作是什么

当前工作簿宏 被自动删除

什么是AutoML？(What is AutoML?)

The development of a model involves a lot of repetitive and tedious tasks inside the Model Development Life Cycle(MDLC), such as tuning the hyper-parameters, generating and selecting features. These tasks consume a lot of time during the development as they are iterative and various permutations and combinations have to be tried to arrive at the best possible combination of model parameters. Different AutoML tools try to automate different parts of the MDLC process. While some try to specifically only work on automating the development of the model, others try to also automate the feature engineering, while some others might be taking into aspects other steps in the lifecycle.

模型的开发在模型开发生命周期( MDLC )内涉及许多重复且繁琐的任务，例如调整超参数，生成和选择特征。 这些任务在开发过程中会耗费大量时间，因为它们是迭代的，必须尝试各种排列和组合以实现模型参数的最佳组合。 不同的AutoML工具尝试使MDLC流程的不同部分自动化。 尽管有些人尝试专门仅使模型开发自动化，而另一些人尝试也使要素工程自动化，而另一些人则可能正在考虑生命周期中的其他步骤。

AutoML tries to achieve its goal by removing such tedious tasks from the hands of the DS which will help the DS and the organization save precious time and money. The approach taken by AutoML tools to achieve this goal, however, varies widely as each of them has different capabilities. While some tools try to automate the entire MDLC, others try to automate certain parts of the Lifecycle.

AutoML试图通过从DS手中除去繁琐的任务来实现其目标，这将帮助DS和组织节省宝贵的时间和金钱。 但是，AutoML工具为实现此目标而采取的方法千差万别，因为每种工具具有不同的功能。 有些工具尝试使整个MDLC自动化，而其他工具则尝试使生命周期的某些部分自动化。

AutoML can, in fact, be described by a phrase — ‘process of automating the automation of automation’; software development is a process used to leverage computing process and thus reduce human hours required for a task. There are certain tasks that can’t normally be performed by computers as it requires a huge set of rules, Machine Learning automates this process. Now, AutoML tries to remove the redundant and repetitive processes inside the MDLC. This makes the above phrase to describe AutoML, quite apt.

实际上，AutoML可以用短语“自动化自动化过程”来描述。 软件开发是一种用于利用计算过程并因此减少任务所需的工时的过程。 有些任务通常无法由计算机执行，因为它需要大量规则，因此机器学习可以自动执行此过程。 现在，AutoML尝试删除MDLC内的冗余和重复进程。 这使得上面描述AutoML的短语非常恰当。

AutoML如何减轻复杂性？ (How does AutoML alleviate the complexities?)

AutoML tools usually have a Graphical User Interface(GUI) which can help guide users without the need for a technical/programming knowledge. This is commonly called the ‘no-code AI’ approach. However, there are also tools which involve the use of code to work with it.

AutoML工具通常具有图形用户界面(GUI)，可以帮助指导用户而无需技术/编程知识。 这通常称为“无代码AI”方法。 但是，也有一些工具涉及使用代码进行处理。

为什么需要AutoML？ (Why is AutoML required?)

Imagine that you have a huge retail company called ABC. ABC has stores throughout the world, of multiple sizes and capacity and is rapidly explained. The following are a few questions it can try to answer using Machine Learning:

想象一下，您有一家名为ABC的大型零售公司。 美国广播公司在世界各地拥有多家商店，规模和容量多种多样，并得到Swift解释。 以下是它可以尝试使用机器学习回答的一些问题：

Supply Chain Optimization

供应链优化

• When to ship a certain product?
  什么时候发货？
• What products could create maximum profit?
  哪些产品可以创造最大利润？
• Which item is most likely to go out of stock in a certain stock at a certain point of time?
  在特定的时间段内，哪种产品最有可能缺货？

Promotion Management

推广管理

• Which type of campaign is the most effective?
  哪种广告系列最有效？
• Are the campaigns having a combined effect or are they independent of each other?
  这些活动是否具有共同的作用，或者它们彼此独立？
• How much money should be spend on promotions in certain areas?
  在某些地区，促销应花费多少钱？
• When should promotions be carried out?
  什么时候进行促销？

Digital Marketing Management

数字营销管理

• Which type of marketing will be the most effective?
  哪种营销方式最有效？
• Will digital marketing campaigns have a significant impact on the area?
  数字营销活动会对该地区产生重大影响吗？

Customer Analysis

客户分析

• What type of customers are most likely to visit a store?
  哪种类型的客户最有可能光顾商店？

The questions listed above are a small fragment of the entire world of applications of Data Science. To implement and maintain these, thousands of Data Science will be required, which, for many companies, might not be feasible. AutoML is hence, a necessary evolution for companies to enhance the productivity of its Data Science divisions.

上面列出的问题只是整个数据科学应用领域的一小部分。 要实施和维护这些技术，将需要成千上万的数据科学，这对于许多公司而言可能不可行。 因此，AutoML是公司提高其数据科学部门生产力的必要演进。

AutoML是否会影响数据科学家的工作机会？ (Will AutoML affect the job opportunities of Data Scientist?)

AutoML tools can only perform specific tasks in the MDLC and is meant as an assistance for the Data Scientist. It is in no way meant to replace the Data Scientist, primarily because, it is just not possible. Data Science is not just the culmination of mathematical and technical skills but also domain knowledge, which is something that cannot be automated. A model development is not just about engineering features and training models but a lot more. These tools might never be able to have the intuition that a human brain has regarding concepts that are of primal importance for a specific business use-case.

AutoML工具只能在MDLC中执行特定任务，并且是对数据科学家的帮助。 这绝不是要替换数据科学家，主要是因为这是不可能的。 数据科学不仅是数学和技术技能的巅峰之作，而且还是领域知识，这是无法自动化的。 模型开发不仅涉及工程功能和训练模型，还涉及更多内容。 这些工具可能永远无法像人脑对特定业务用例至关重要的概念具有直觉。

It is important to understand that AutoML is specifically called AutoML(Automated Machine Learning) and not AutoDS(Automated Data Science)

重要的是要理解AutoML被专门称为AutoML(自动机器学习)而不是AutoDS(自动数据科学)

After a model has been developed, there are still numerous aspects that have to be taken into consideration such as its compliance to the legal structure of a geographical region, racial bias and its decision making process. Imagine that you have a model deployed to filter out credit card applications and a customer asks you why his/her credit card application was rejected. As a company, under the GDPR, you are liable to explain the process and why the decision was made. Hence, it is very clear that AutoML tools cannot be a replacement to Data Scientists and can only act as an assistant.

建立模型后，仍然需要考虑许多方面，例如其对地理区域法律结构的遵守，种族偏见及其决策过程。 想象一下，您已经部署了一个模型来筛选出信用卡申请，并且客户问您为什么拒绝他/她的信用卡申请。 作为公司，根据GDPR，您有责任说明流程以及做出该决定的原因。 因此，很明显，AutoML工具不能替代数据科学家，而只能充当助手。

翻译自: https://towardsdatascience.com/will-automl-take-away-my-job-what-is-it-34a2d01f6848

当前工作簿宏 被自动删除