tableau数据案例_Tableau对数据科学家有用吗？

tableau数据案例

目录 (Table of Contents)

1. Introduction
   介绍
2. Tableau
   画面
3. What Could Improve
   有什么可以改善的
4. The Great
   最棒的
5. Summary
   摘要

介绍 (Introduction)

While it is not completely necessary to have Tableau as a part of your skillset, it can still provide to be useful in your day-to-day as a Data Scientist. Having worked at a few companies (business, finance, and tech), I can say that some companies completely do not use Tableau, especially the Data Scientists working there, while some companies can expect to have their Data Scientists and Machine Learning Engineers working with Tableau a few days per week. Ultimately, it depends on you, your team, and your business, if you want to utilize Tableau. Just because it is not already implemented in your current process, does not mean it cannot still be added whenever it feels like it is necessary to use it. Therefore, I am going to highlight what Tabeau could improve on, and what it is great at — in terms of Data Science and Machine Learning. Tableau as a Data Analtyics and visualization tool is outstanding and I recommend it. For Data Scientists, there are pros and cons and I will be describing those below.

虽然不一定完全需要Tableau作为技能组合的一部分，但它仍然可以对您作为数据科学家的日常工作有用。 我曾在一些公司( 商业，金融和技术公司 )工作过，我可以说有些公司完全不使用Tableau，尤其是在那里工作的数据科学家，而有些公司可以期望他们的数据科学家和机器学习工程师可以与他们合作Tableau每周几天。 最终，如果要使用Tableau，它取决于您，您的团队和您的业务。 仅仅因为它尚未在您当前的过程中实现，并不意味着在感觉需要使用它时仍不能添加它。 因此，在数据科学和机器学习方面，我将重点介绍Tabeau可以改进的方面以及它的擅长之处。 Tableau作为数据分析和可视化工具非常出色，我建议您使用它。 对于数据科学家来说，有优点也有缺点，我将在下面进行描述。

画面 (Tableau)

Tableau [2], is a useful tool, primarily for Business and Data Analysts. Some positions at companies even have a designated Tableau Developer that will only focus on creating reports and dashboards for their respective audiences or stakeholders. Some of the important ways that you can “change the way you think about data” with Tableau are the following:

Tableau [2]是有用的工具，主要用于业务和数据分析师。 公司的某些职位甚至都有指定的Tableau开发人员，该开发人员仅专注于为其各自的受众或利益相关者创建报告和仪表板。 使用Tableau可以“ 改变对数据的看法 ”的一些重要方法如下：

• Fast Analytics
  快速分析
• Ease of Use
  使用方便
• Big Data, Any Data
  大数据，任何数据
• Smart Dashboards
  智能仪表板
• Update Automatically
  自动更新
• Share in Seconds
  秒分享

I love using Tableau, personally, when creating visualizations quickly. I can use fancier Python packages, but sometimes, utilizing Tableau’s SQL database connection services allows you to query, and essentially drag and drop to describe or visualize your data so that you can tell your story with style. Here is the link to the webpage for Tableau:

快速创建可视化效果时，我个人喜欢使用Tableau。 我可以使用更高级的Python程序包，但是有时，利用TableauSQL数据库连接服务可以查询，并且实质上是拖放操作来描述或可视化数据，以便您可以用风格讲述故事。 这是Tableau网页的链接：

有什么可以改善的 (What Could Improve)

Here’s what Tableau doesn’t do well for Data Scientists.

下面是的Tableau 没有 为数据科学家做的很好。

Keep in mind, I am not saying this for general Data Analysts, what Tableau is intended for, I am highlighting these points for Data Scientists. These are three points that come to mind when working with Tableau — as you will see later on, there are plenty more positives or pros than possible improvements to be made:

请记住，我不是要对一般的数据分析师说这话，Tableau打算这样做，而是要为数据科学家强调这些要点。 使用Tableau时，请牢记以下三点-正如您稍后将看到的，比起可能要改进的地方，有很多积极或有利的方面：

• Cannot integrate with a Jupyter Notebook

  无法与Jupyter Notebook集成

As Data Scientists, integration and automation are key. You like to or are used to having a process that unites all your processes together so that when you perform your business problem use case, exploratory data analysis, feature engineering, model building, and deployment, you can easily refer to these steps all in one place or all in one process that is connected. I would enjoy having a way to display the visualizations made in Tableau to be shown in a Jupyter Notebook, or some type of similar integration.

作为数据科学家，集成和自动化是关键。 您喜欢或习惯于将所有流程组合在一起的流程，因此在执行业务问题用例，探索性数据分析，功能工程，模型构建和部署时，可以轻松地将所有这些步骤一并参考将一个或多个连接在一起。 我希望有一种方法可以显示在Tableau中制作的可视化内容，以显示在Jupyter笔记本中，或某种类型的类似集成中。

However, the fact that you can have a live database connection for direct SQL querying and report generation in Tableau is both awesome and useful.

但是，您可以使用实时数据库连接来直接在Tableau中进行SQL查询和生成报表这一事实既很棒又有用。

• Can be slow sometimes

  有时可能很慢

Now, this point may not happen to you at all, but sometimes you can find yourself with several tabs or sheets and dashboards, and all of a sudden you have this giant Tableau workbook that freezes, and is somewhat frustrating to keep on making new dashboards without deleting old ones.

现在，这一点可能根本就没有发生，但是有时您会发现自己有几个选项卡，工作表和仪表板，突然之间，您的Tableau工作簿冻结了，使继续制作新的仪表板有些沮丧而不删除旧的。

As a work-around, sometimes I start with a sample dataset and make the dashboards as a proof of concept, and then will apply the whole dataset for the final form.

解决方法是，有时我从示例数据集开始，将仪表盘作为概念证明，然后将整个数据集应用于最终表单。

• Has a limit to how many Data Science applications

  对多少个数据科学应用程序有限制

This point is not necessarily bad, since Tableau is not a Data Science tool. It does have some awesome applications that I will discuss below, but it would be interesting to see a separate Data Science section in the future.

这一点不一定很糟糕，因为Tableau并不是数据科学工具。 它的确有一些很棒的应用程序，我将在下面进行讨论，但是将来看到单独的数据科学部分会很有趣。

最棒的 (The Great)

Here’s what Tableau does do well for Data Scientists.

下面是 它 的Tableau 数据科学家做的很好。

• Visualizes datasets well for exploratory data analysis (EDA)

  很好地可视化数据集以进行探索性数据分析(EDA)

EDA is often overlooked in Data Science processes, and it can make or break your model. Having the ability to visualize your data quickly before building the model (not having to write any Python code), is extremely beneficial. It is also useful to display charts, graphs, or other forms of visualizations for your Data Science or Machine Learning model metrics (like average accuracy per day).

EDA通常在数据科学流程中被忽略，它可以建立或破坏您的模型。 具有在构建模型之前快速可视化数据的能力( 无需编写任何Python代码 )，这是非常有益的。 为数据科学或机器学习模型指标( 例如每天平均准确性 )显示图表，图形或其他形式的可视化效果也很有用。

A neat feature is that you can set up alerts if the value of the graphed data is below or above a certain threshold — say you want to get an email that alerts you that your model just went under 80% accuracy for the first time in months, this alert could then be investigated where before it would be overlooked.

一项巧妙的功能是，如果图形数据的值低于或高于特定阈值，则可以设置警报-假设您想收到一封电子邮件，提醒您您的模型几个月来首次准确性低于80％ ，然后可以在忽略该警报之前对其进行调查。

• In general, is a step up from Matplotlib and Seaborn Python libraries

  总的来说，是从Matplotlib和Seaborn Python库升级的

Sometimes it takes a lot of work, or a lot of Python code (I do not use R so I will not speak to that), to create a somewhat unappealing chart. With Tableau, you can make fancy visualizations in a few seconds without coding.

有时需要花费大量的工作或大量的Python代码( 我不使用R，所以我不再赘述 )来创建一个不太吸引人的图表。 使用Tableau，您可以在几秒钟内完成精美的可视化效果而无需编写代码。

• Visualizes summary success metrics for data science models well

  很好地可视化数据科学模型的摘要成功指标

I spoke to this point a little above, but want to stress that you can visualize your model metrics or output easily with Tableau — assuming your results, confidence score, suggestions, etc, are stored in a SQL database. You can, perhaps, output all of the categorizations from your model that had a low confidence score so that they can be manually reviewed by subject matter experts in your company, furthering your accuracy improvements along the way.

我在上面稍微提到了这一点，但想强调一下，您可以使用Tableau可视化模型指标或轻松输出-假设结果，置信度得分，建议等存储在SQL数据库中。 也许您可以从模型中输出置信度得分较低的所有分类，以便公司的主题专家可以手动检查它们，从而进一步提高准确性。

• Integrates well with SQL queries well

  很好地与SQL查询集成

What you can do in SQL, you can do in Tableau, essentially. You can paste your queries and reference them to make anything in Tableau — you can also use a static Excel/CSV file, for example, if you are working with data that does not necessarily need to be live.

从本质上讲，您可以使用SQL进行操作，也可以使用Tableau进行操作。 您可以粘贴查询并引用它们以在Tableau中进行任何操作-例如，如果要处理不一定需要实时处理的数据，还可以使用静态Excel / CSV文件。

• You can do clustering!

  您可以进行群集！

k-means algorithm with Tableau!

Tableau的k均值算法！

I saved the best for last, this benefit of Tableau is awesome. You can perform a clustering model [3 ] without any code! Yes, since you are not building it yourself, it will not be as tunable,

我把最好的留到了最后，Tableau的好处非常棒。 您无需任何代码即可执行聚类模型 [3]！ 是的，由于您不是自己构建它，所以它不会那么可调，

…but you can make a Data Science model in less than a day with this awesome feature of Tableau.

…但是，借助Tableau的强大功能，您可以在不到一天的时间内建立数据科学模型。

A great use case for clustering in Tableau is a quick and easy way to find similarities between groups of customers so that you can market towards them differently — think marketing campaigns.

在Tableau中进行群集的一个很好的用例是一种快速简便的方法来查找客户组之间的相似性，以便您可以针对他们进行不同的市场营销-考虑营销活动。

摘要 (Summary)

Photo by Kevin Ku on Unsplash [4].

凯文·库 ( Kevin Ku)在 Unsplash [4]上的照片。

Ultimately, Tableau was not intended for Data Scientists, so it is rather impressive how useful it is for us. There are several more pros than cons that I have discussed, and most likely even more pros with some creativity. As someone who is a visual learner and presenter, I really enjoy using Tableau. It is a nice break from Python code for me, and I appreciate that the visualizations can be as easy as dragging and dropping columns from your data.

最终，Tableau并非旨在供数据科学家使用，因此它对我们的实用性令人印象深刻。 我讨论过的利弊多于利弊，并且更有可能具有创造力的利弊更多。 作为视觉学习者和演示者，我真的很喜欢使用Tableau。 对我而言，这是与Python代码的一个不错的突破，并且我欣赏可视化效果就像从数据中拖放列一样容易。

I recommend using Tableau, and yes, it is useful for Data Scientists.

我建议使用Tableau，是的，它对数据科学家很有用。

Thank you for reading my article. I hope you found it useful. If you enjoyed it, please let me know, and if you have any recommendations or comments, feel free to submit those down below. Thank you!

感谢您阅读我的文章。 希望你觉得它有用。 如果您喜欢它，请让我知道，如果您有任何建议或意见，请随时在下面提交。 谢谢！

翻译自: https://towardsdatascience.com/is-tableau-useful-for-data-scientists-46d355a14b62

tableau数据案例