數據可視化機器學習工具在線

重點 (Top highlight)

There’s no scarcity of posts online about ‘fancy’ data topics like data modelling and data engineering. But I’ve noticed their cousin, data visualization, barely gets the same amount of attention. Among data practitioners in my field, I find there is solid consensus that data viz is an important skill that’s worth devoting time to learn. However, we somehow spend more time learning complex models over figuring out why pie charts are no-nos.

在線上沒有關于“花式”數據主題(如數據建模和數據工程)的帖子。 但是我注意到他們的表弟，數據可視化，幾乎沒有受到同樣的關注。 在我所在領域的數據從業者中，我發現有一個牢固的共識，即數據可視化是一項重要的技能，值得花時間學習。 但是，我們花了更多的時間來學習復雜的模型，而不是弄清楚為什么餅圖不行。

When I was just starting out in my career, the hype was really in modelling. I considered making graphs and visualizations ‘chores’, and I thought that the depth of my fancy data science knowledge was the greatest determinant of how much value I could bring.

當我剛開始我的職業生涯時，炒作真的是在建模。 我考慮過使圖表和可視化成為“瑣事”，并且我認為我對數據科學知識的深度決定了我可以帶來多少價值。

As I developed the data viz aspect of my skill set, I picked up valuable lessons that now influence how I approach everything that I do. These lessons have proven extremely useful for me in my career journey, and I’d like to build the case here for why data viz is a core tool in any data person’s skill set.

在開發技能的數據可視化方面時，我吸取了寶貴的經驗教訓，這些經驗教訓現在影響著我處理工作方式的方式。 在我的職業生涯中，這些課程對我來說非常有用，我想在這里舉例說明為什么數據可視化是任何數據人員技能中的核心工具。

學習1：如果您希望人們做正確的事，則必須使其變得容易 (Learning 1: If you want people to do the right thing, you have to make it easy)

A common frustration I hear from data people is that business stakeholders often don’t seem to make decisions that align with what data analysts have found to be optimal. This leads to a lose-lose scenario where stakeholders don’t get the results they’re looking for, while analysts get frustrated with ‘wasted work.’ I was also once stuck in the ‘if only stakeholders listened’ mentality.

我從數據人員那里聽到的一個普遍沮喪是，業務利益相關者似乎常常沒有做出與數據分析師認為最佳的決策一致的決策。 這導致了失敗的情況，即利益相關者沒有得到他們想要的結果，而分析師則對“浪費的工作”感到沮喪。 我也曾經陷入“如果只有利益相關者傾聽”的心態。

Darkhorseanalytics.Darkhorseanalytics 。

It was working with User Experience (UX) Product Designers (and surprisingly, not a data seminar) that brought me to rethink this mindset. They brought attention to things I would have called ‘trivial’ before, critiquing how websites and applications were built: ‘this banner is too big’, ‘this button should be colored blue’, etc — and for good reason. They told me that ‘If we don’t do X, the user will have trouble doing what we want them to do.’

它與用戶體驗(UX)產品設計師(而且令人驚訝的是，不是數據研討會)合作，使我重新考慮了這種思維方式。 他們引起了我以前所謂的“瑣碎”的注意，并批評了網站和應用程序的構建方式：“此橫幅太大”，“此按鈕應被涂成藍色”等，這是有充分理由的。 他們告訴我， “如果我們不做X，用戶將很難做我們想要他們做的事情。”

Coming back to the context of data, this insight is exactly why I would argue that data visualization is powerful. Well thought-out visuals, rather than ugly charts, most clearly explain to stakeholders why a data-backed recommendation will be most beneficial. After all, the main role of a data professional is not just to churn out models and analysis, but also to inspire data-driven action.

回到數據上下文，這種見解正是我認為數據可視化功能強大的原因。 經過深思熟慮的視覺效果，而不是難看的圖表，可以最清晰地向利益相關者解釋為什么數據支持的建議將最有益。 畢竟，數據專業人員的主要作用不僅在于建立模型和分析，還在于激發數據驅動的行動。

As a data professional (or any profession really), maybe we should think twice the next time ‘slides are a chore’ crosses our minds.

作為數據專業人士(或者實際上是任何專業)，也許我們下次應該想到“幻燈片是一件瑣事”時就要三思而行。

學習2：一張圖片值一千字一分鐘。 (Learning 2: A picture is worth a thousand words AND a thousand minutes.)

tjpalanca.comtjpalanca.com

Full disclosure — I got irritated whenever people asked me to edit what seemed like irrelevant details in presentations. ‘Change the y-axis label from revenue to Revenue’, ‘Make this category green instead of red’, and ‘Add another graph,’ were common phrases I dealt with in my everyday work. I guess I did not like hearing them because I thought they’d mean that I’d have to re-render a report and waste time on a task I didn’t care about.

完全公開-每當有人要求我編輯演示文稿中似乎無關緊要的細節時，我都會很生氣。 我在日常工作中經常使用“將y軸標簽從“收入”更改為“收入””，“將此類別設置為綠色而不是紅色”和“添加其他圖表”。 我想我不喜歡聽到他們的聲音，因為我認為他們的意思是我必須重新提交報告，并在我不關心的任務上浪費時間。

This was admittedly ironic for me, since when doing research, I [hated] reading papers that were written as if it was the author’s goal to sound as complex as possible. I’d have to spend hours dissecting a set of paragraphs and equations in order to comprehend a piece of insight.

誠然，這對我具有諷刺意味，因為在進行研究時，我[討厭]閱讀所寫的論文，好像作者的目標是聽起來盡可能的復雜。 為了理解這一點，我不得不花幾個小時來剖析一組段落和等式。

On the other hand, what I find most enjoyable and engaging to read are papers with extensive visualisations. In those papers, it was clear that the author took time and effort to create data visuals to aid a reader in understanding the point he or she wanted to get across.

另一方面，我覺得最有趣，最吸引人的是那些具有廣泛可視化效果的論文。 在這些論文中，很明顯，作者花了時間和精力來創建數據視覺效果，以幫助讀者理解他或她想傳達的觀點。

Even while trying to understand complex topics in other mediums, I can’t count how many hours I’ve saved thanks to Youtube channels like 3Blue1Brown or Kurzgesagt that pair rich explanations with engaging visuals to illustrate their point.

即使嘗試理解其他媒體中的復雜主題，由于YouTube頻道(如3Blue1Brown或Kurzgesagt)將豐富的說明與引人入勝的視覺效果結合起來，我無法節省多少時間。

演示地址

3Blue1Brown on Neural Networks神經網絡上的3Blue1Brown

These materials take time to make, and in analytics, making visuals can sometimes appear as optional. However, that 10 minutes you spent making a visual is 5–10 minutes saved for EACH individual trying to understand your point. This especially gets important in a fast-moving business, where decisions have to be made on the fly and a simple visual would have helped push through blockers to get to the action.

這些材料制作需要花費時間，在分析中，制作視覺效果有時可能是可選的。 但是，對于每一個試圖理解您的觀點的人來說，您花10分鐘的時間進行視覺處理可以節省5-10分鐘。 在快速發展的業務中，這尤其重要，因為該業務必須即時做出決策，而簡單的視覺效果將有助于推動阻止者采取行動。

It can’t be overstated how the effort to make good data visualisations pays in the long run by influencing organizations and contributing to the efficiency of the decision-making process.

從長遠來看，通過影響組織并為決策流程的效率做出貢獻，使良好的數據可視化所付出的努力是如何付出的，這是不夸張的。

學習3：解決問題并不一定是一項艱巨的任務！ (Learning 3: Problem solving does not need to be a solitary undertaking!)

I found this tweet by Mat Velloso very funny and relatable:

我發現Mat Velloso的這條推文非常有趣且相關：

演示地址

I assume the reason this happens is because companies find it difficult to measure the level of complexity of the problems they’re facing and “AI” sounds like a good catch-all solution. Even as a data practitioner, I’ve also been guilty of this, falling into analysis-paralysis and trying every single model, hoping one will magically solve everything. This is a big waste of the analyst’s (and the stakeholder’s) time.

我認為發生這種情況的原因是，公司發現難以衡量所面臨問題的復雜程度，而“ AI”聽起來像是一個萬能的解決方案。 即使作為一名數據從業者，我也對此感到內,，陷入了分析癱瘓并嘗試每個模型，希望一個模型能神奇地解決所有問題。 這浪費了分析師(和利益相關者)的時間。

It’s funny and all, but it goes to show that we can all still work on how we collectively understand complex data problems.

一切都很好笑，但這表明我們仍然可以共同致力于理解復雜數據問題。

Notice the number of reactions in this post in our Slack Channels (I blurred some details):

請注意我們的Slack頻道中此帖子的React數量(我模糊了一些細節)：

On the other hand, check out the reactions (and replies!) on this other post with a data viz attached:

另一方面，請查看此帖子的React(和回復！)，并附加數據：

The posts show how people are happy to share thoughts and insights, provided they can relate and easily understand the context.

帖子顯示了人們如何樂于分享思想和見解，只要他們能夠聯系在一起并 輕松了解上下文。

Modelling is indeed useful when we are sure that it is the most efficient solution to address our problem. However, the beauty about data viz is the communication piece that allows us to make problem solving a collaborative experience. After all, there is no scarcity of ideas from subject-matter experts whose only barrier is the technical data aspect. Imagine, if data viz made the problem scope clearer for other people, those who see our visuals can start ideating solutions on a business problem and the ripple effect amplifies. Suddenly from one person making a model (you), other people are already giving you ideas to further improve your work (and possibly even a solution that doesn’t need a model!)

當我們確定建模是解決我們的問題的最有效解決方案時，建模確實有用。 但是，關于數據即可視化的美是使我們能夠解決協作體驗問題的溝通工具。 畢竟，主題專家們并不缺乏想法，他們的唯一障礙是技術數據方面。 想象一下，如果數據可視化使其他人更清楚地了解問題的范圍，那么看到我們的視覺效果的人就可以開始就業務問題提出解決方案，并且連鎖React會不斷擴大。 突然間，一個人(您)在制作模型，其他人已經在為您提供構想，以進一步改進您的工作(甚至可能不需要模型的解決方案！)

學習4：有一個時間和地點，一切都可視 (Learning 4: There is a time and place AND VISUAL for everything)

演示地址

Early in my career, my development goals were bullet points of technical topics that I wanted to learn and apply: recommendation systems, image processing, bayesian modelling, etc. So of course, I hopped on the first chance to apply and try out these techniques, thinking that these will generate the greatest amount of value with my limited amount of time.

在我職業生涯的早期，我的發展目標是我想學習和應用的技術主題的要點：推薦系統，圖像處理，貝葉斯建模等。因此，我當然希望有機會應用和嘗試這些技術，認為這些將在有限的時間內產生最大的價值。

Most of the time, however, a visual can offer the best bang-for-buck. In the model development process after all, there is always the exploration piece which usually involves data viz. In this step, we can already get a lot of insights from different visuals:

但是，大多數情況下，視覺效果可以提供最佳的性價比。 畢竟，在模型開發過程中，總是存在通常涉及數據的探索部分。 在這一步中，我們已經可以從不同的視覺效果中獲得很多見解：

• Dodged Bar Graph — Which brand of milk tea sold the most?
  閃避條形圖—哪個品牌的奶茶銷量最高？
• Stacked Bar Graph — Which milk tea branch contributes the most to revenue?
  堆積條形圖—哪個奶茶分支對收入的貢獻最大？
• Line Graph — Did average cost of milk tea decrease/increase?
  線形圖—奶茶的平均成本是否降低/增加了？
• Scatter Plot — How is age related to number of milk tea purchased?
  散點圖—年齡與購買的奶茶數量有何關系？
• Density Plot — What is the distribution of the age of milk tea drinkers?
  密度圖-喝奶茶的年齡分布如何？
• Tables — What is the total, average, mean, and median of milk tea purchases per customer across all months?
  表格-在每個月中，每位客戶購買奶茶的總數，平均值，平均值和中位數是多少？
• Pie Chart — No.
  餅圖-不

The Extreme Presentation Method by Dr. Abela的極端演示方法重現視覺

Data viz helps us dissect problems into comprehensible pieces, and potentially equips us to answer business questions without the need to apply complex methodologies. Going deep into visualisation will force you to ask “Are we looking at the right things?” and eventually — “Are we being led to the best strategies?”

數據可視化幫助我們將問題分解為可理解的部分，并有可能使我們無需使用復雜的方法即可回答業務問題。 深入可視化將迫使您問“我們在看正確的東西嗎？” 最終-“我們被引導到最佳策略了嗎？”

Data Viz是一種同時培養軟技能和硬技能的實踐 (Data Viz is a practice that cultivates both soft and hard skills at the same time)

• Hard Skills: Data Wrangling (How do we transform data into its pre-graph form?), Grammar of Graphics (How are visualizations constructed?)

  硬技能：數據整理(如何將數據轉換為圖形形式)， 圖形語法 (如何構建可視化？)

• Soft Skills: Communication (What visual will effectively drive home a point?), Critical Thinking (What visual will best convey a good strategy?)

  軟技能：交流(哪種視覺方法可以有效地傳達觀點？)，批判性思維(哪種視覺方法可以最好地傳達一種良好的策略？)

From these experiences, I believe learning the principles behind data visualisation is beneficial for any data practitioner, and possibly even for people from other professions as well. Data viz is how we can communicate the complexity of data to the visual learners of the world (65% of the population, according to Google). We cannot skip learning this extremely useful and practical skill, and I’ll wait for the day when pie charts only exist in the form of memes.

從這些經驗中，我相信學習數據可視化背后的原理對任何數據從業者都是有益的，甚至對于其他專業人士也可能是有益的。 數據即是我們如何將數據的復雜性傳達給世界范圍內的視覺學習者(根據Google的數據，占總人口的65％)。 我們不能跳過學習這一極其有用和實用的技能，而我將等待餅圖僅以模因形式存在的那一天。

演示地址

Geckoboard on Pie ChartsGeckoboard在餅圖上的短短一分鐘視頻

This post is mostly a compilation of opinions I formed in my data career. I’m super open to thoughts and comments from fellow practitioners and aspiring ones alike! Please feel free to shoot an email to nfrimando@gmail.com or connect with me via LinkedIn. Shout out to Fernandina Ko who helped me edit this piece, Christian San Jose for inviting me to the UX Fundamentals Class, and TJ Palanca for the minimalist map visual!

這篇文章大部分是我在數據職業生涯中形成的觀點的匯編。 我非常歡迎同修和有抱負的人的想法和評論！ 請隨時發送電子郵件至nfrimando@gmail.com或通過LinkedIn與我聯系。 向Fernandina Ko大喊大叫 ，后者幫助我編輯了這篇文章， Christian San Jose邀請我參加UX基礎知識課程，向TJ Palanca提供了極簡地圖視覺效果！

翻譯自: https://medium.com/swlh/why-you-cant-skip-learning-data-visualization-6314896ccdc0

數據可視化機器學習工具在線

總結

以上是生活随笔為你收集整理的数据可视化机器学习工具在线_为什么您不能跳过学习数据可视化的全部內容，希望文章能夠幫你解決所遇到的問題。

如果覺得生活随笔網站內容還不錯，歡迎將生活随笔推薦給好友。