什么是消費者行為？ (What is Consumer Behavior?)

consumer behavior is the study of individuals, groups, or organizations and all the activities associated with the purchase, use, and disposal of goods and services, and how the consumer’s emotions, attitudes, and preferences affect buying behavior.

消費者行為是對個人，團體或組織以及與購買，使用和處置商品和服務相關的所有活動的研究，以及消費者的情感，態度和偏好如何影響購買行為。

什么影響消費者行為？ (What affects consumer behavior?)

Making decisions are very dynamic processing and a lot of things could affect consumer behavior. Analyzing consumer behavior data are allowed us to present products or services in a way that generates a maximum impact on consumers. Consumer behavior is often influenced by different factors such as:

決策是非常動態的處理，很多事情都會影響消費者的行為。 分析消費者行為數據使我們能夠以對消費者產生最大影響的方式展示產品或服務。 消費者行為通常受不同因素的影響，例如：

• Personal factors: a specific characteristic such as age, race, gender, culture, income, personal habit/interest, etc
  個人因素：特定特征，例如年齡，種族，性別，文化，收入，個人習慣/興趣等
• Psychological factors: an individual’s response to a marketing message will depend on their perceptions and attitudes.
  心理因素：個人對營銷信息的React將取決于他們的看法和態度。
• Company Website: web design, products’ reviews from other users, recommendation System
  公司網站：網頁設計，其他用戶的產品評論，推薦系統
• Company’s physical store service
  公司的實體店服務
• Social: Policy, Government, Economy, Competitors, World of mouth, Friends, and Family, etc
  社會：政策，政府，經濟，競爭對手，口碑，朋友和家人等
• Advertising: Email, Mail, Magazine, Cookie, Social Media, Affiliate(Cashback Website), Partnership (Events like NBA, World Cup, TV Shows)
  廣告：電子郵件，郵件，雜志，Cookie，社交媒體，會員(現金返還網站)，合作伙伴關系(NBA，世界杯，電視節目之類的活動)

為什么消費者行為數據很重要？ (Why Consumer Behavior data is important?)

Understanding consumer behavior is a vital aspect of marketing. Based on consumer behavior data, we are able to know how consumers make decisions and how potential customers will respond to new products or new services. It is important to explore actionable insights from the data to support companies to put forward corresponding strategies.

了解消費者行為是營銷的重要方面。 根據消費者行為數據，我們能夠知道消費者如何做出決策以及潛在客戶將如何響應新產品或新服務。 重要的是，從數據中探索可行的見解，以支持公司提出相應的策略。

For example, if you visit this kind of website, you do not know where should you start with and you will lose your patient then you will want to leave this website immediately. That is why we do the tests (such as AB test, usability test, etc) to gain insights into customer behavior in order to optimize the customer journey and improve key KPIs — Conversion Rate, Revenue, Customer Life Time Value, and AOV (Average Order Value).

例如，如果您訪問這種網站，則不知道從哪里開始，您會失去耐心，那么您將希望立即離開該網站。 這就是為什么我們進行測試(例如AB測試，可用性測試等)以深入了解客戶行為，從而優化客戶旅程并改善關鍵KPI(轉化率，收入，客戶生命周期價值和AOV(平均))的原因訂單價值)。

A good and nice e-commerce website can greatly affect consumer behavior.

一個好的電子商務網站可以極大地影響消費者的行為。

行為經濟學的視角 (The Perspective of Behavioral Economics)

Behavioral Economics is the study of psychology as it relates to the economic decision-making processes of individuals and institutions. It could be combined with consumer behavior to study how people make decisions and how to affect people to make decisions with ‘invisible hands’.

行為經濟學是對心理學的研究，因為它涉及個人和機構的經濟決策過程。 它可以與消費者行為相結合，研究人們如何做出決策以及如何影響人們用“看不見的手”做出決策。

Dan Ariely is an Israeli-American professor of psychology and behavioral economics at Duke University. He mainly studies the role of human psychological phenomena in economics and how to use human psychological phenomena to influence people’s economic behavior. There are some classical examples to show how to influence consumers’ behavior.

丹·阿里利(Dan Ariely)是杜克大學(Duke University)的一名美籍美國裔心理學和行為經濟學教授。 他主要研究人類心理現象在經濟學中的作用以及如何利用人類心理現象影響人們的經濟行為。 有一些經典的例子來說明如何影響消費者的行為。

1st example: When you apply for driver’s license at DMV you can see there is a part in your application to show if you wish to donate your organs and tissues after declared death. Dan Ariely did research about the ratio of the people who are willing to donate their organs in European countries and here is the bar chart for the result.

1日例如：當你在DMV申請駕駛證，你可以看到有在應用程序中顯示的一部分，如果你想宣布死后捐出你的器官和組織。 丹·阿里利(Dan Ariely)研究了愿意在歐洲國家捐獻器官的人口比例，這是得出結果的條形圖。

The first four countries' situations are very different from the remaining countries. You might think those differences due to cultural and religious reasons, however, the cultures and religions of some countries are very similar. After research, he finally found that the biggest difference between those countries is the DMV application designs.

前四個國家的情況與其余國家大不相同。 您可能會認為這些差異是由于文化和宗教原因造成的，但是某些國家的文化和宗教非常相似。 經過研究，他最終發現這兩個國家之間最大的不同是DMV應用程序設計。

• design1 - Countries with low ratio: “If you wish to donate your organs and tissue, please sign here…”
  design1-比率較低的國家：“如果您想捐獻器官和組織，請在此處簽名……”
• design2 -Countries with high ratio: “If you do not wish to donate your organs and tissue, please sign here…”
  design2-高比率國家：“如果您不希望捐贈器官和組織，請在此處簽名……”

As long as people notice that there is something relative to “organ donation”, by the “default” most people will not sign for design 1 or design 2. So the people’s decision was affected by the different designs of organ donation consent. This is a kind of dependence on the default options because it does not give people psychological pressure.

只要人們注意到與“器官捐贈”有關的事情， 默認情況下，大多數人就不會簽署外觀設計1或外觀設計2的簽名。因此，人們的決定會受到不同形式的器官捐贈同意設計的影響。 這是對默認選項的一種依賴，因為它不會給人們帶來心理壓力。

2nd example: If one brand always offers a “big sale” promotion, most consumers will think that the real price is cheaper than the after-sale price and might also think “low price = low quality”. On the other hand, another expensive brand rarely offers discounts. If there is a big sale of this brand, most consumers are more likely to purchase and maybe buy some products that they might not really need.

第二例子：如果一個品牌總是提供了一個“大甩賣”的促銷活動，大多數消費者會認為真正的價格比售后價格更便宜，也可能會想“低價格=低品質”。 另一方面，另一個昂貴的品牌很少提供折扣。 如果該品牌的銷售量很大，那么大多數消費者更有可能購買甚至購買一些他們可能真正不需要的產品。

3rd example: There is a traveling advertisement which is for a 7-DAY tour for Rome or Paris. People have two equal options — $1800 includes the fees of hotels and meals for both different cities. The number of people who chose Roma is similar to that of people who chose Paris.

第三例子：有一個移動廣告是一個7日游羅馬或巴黎。 人們有兩個平等的選擇-1800美元包括兩個城市的酒店和伙食費。 選擇羅姆人的人數與選擇巴黎人的人數相似。

After a new option was added — ‘Rome: Hotel+No Meal — $1800’, most people chose to go to Rome with the “Hotel+Meal”. Actually adding the “Hotel+No Meal” option is meaningless and obviously not a lot of people will choose it. However, the purpose of the new option is to influence people’s decisions.

在增加了一個新選項“羅馬：酒店+無餐-1800美元”之后，大多數人選擇了帶“酒店+餐”的羅馬。 實際上，添加“酒店+無餐”選項是沒有意義的，并且顯然不會有很多人會選擇它。 但是，新選項的目的是影響人們的決策。

There is an interesting hidden theory behind those examples- Game Theory! Game Theory is a study of strategic interaction among rational decision-makers using mathematical models. If we are able to understand and analyze Consumer Behavior with Behavioral Economics/Game Theory, we can predict people’s behavior and their expectation, and then we can put forward targeted strategies that can affect consumer’s decisions and maximize their’s utility. If you are interested in Game Theory, you can visit my previous blog.

這些示例背后有一個有趣的隱藏理論，即博弈論！ 博弈論是使用數學模型研究理性決策者之間的戰略互動。 如果我們能夠使用行為經濟學/博弈論來理解和分析消費者行為，那么我們就可以預測人們的行為及其期望，然后我們可以提出有針對性的策略來影響消費者的決策并最大化他們的效用。 如果您對博弈論感興趣，可以訪問我以前的博客 。

項目介紹 (Project Introduction)

The goal of this project is to predict what kind of consumers are more likely to click the ad.

該項目的目的是預測哪種類型的消費者更有可能點擊廣告。

The dataset created by Jose Portilla and Pierian Data for his Udemy Course (Python for Data Science and Machine Learning Bootcamp). You can get the data from Kaggle. The data contains ten different columns:

Jose Portilla和Pierian Data為他的Udemy課程(Python for Data Science and Machine Learning Bootcamp)創建的數據集。 您可以從Kaggle獲取數據。 數據包含十個不同的列：

1. Daily Time Spent on a Site — Time spent by the user on a site in minutes.
   網站上花費的每日時間-用戶在網站上花費的時間，以分鐘為單位。
2. Age — Customer’s age in terms of years.
   年齡-客戶的年齡(以年為單位)。
3. Area Income — Average income of geographical area of consumer.
   地區收入-消費者地理區域的平均收入。
4. Daily Internet Usage — Avgerage minutes in a day consumer is on the internet.
   每天的互聯網使用情況-一天的平均消費分鐘數是在互聯網上。
5. Ad Topic Line — Headline of the advertisement.
   廣告主題行-廣告標題。
6. City — City of the consumer.
   城市-消費者的城市。
7. Male — Whether or not a consumer was male.
   男性-消費者是否為男性。
8. Country — Country of the consumer.
   國家-消費者所在的國家。
9. Timestamp — Time at which the user clicked on an Ad or the closed window.
   時間戳記-用戶單擊廣告或關閉的窗口的時間。
10. Clicked on Ad — 0 or 1 is indicated clicking on an Ad or not — Class 0 — not clicked, and Class 1 — clicked.
    單擊廣告-0或1表示是否單擊廣告-0級-未單擊，1級-已單擊。

EDA (EDA)

數值數據與目標 (Numerical Data VS Target)

Based on the image below, we can see that people who spend more time — around 80 minutes on the site are not likely to click the ad, and people who spend around 50 minutes are more likely to click the ad. The average age of the people who clicked the ad is around 40 and the average area income of consumers in class 1 is around 50000. Last, consumers in Class 1 have less daily internet usage. Those subplots are very informative and bring us some basic ideas. Besides numerical data, text data can also bring some useful information. For example, what kind of headline or topic of the ad is more attractive and consumers are more likely to click it.

根據下圖，我們可以看到花費更多時間的用戶-在網站上花費大約80分鐘的時間不太可能點擊廣告，而花費大約50分鐘的用戶則更有可能點擊廣告。 點擊廣告的人的平均年齡為40歲左右，類別1的消費者的平均地區收入約為50000。最后，類別1的消費者的每日互聯網使用量較少。 這些子圖非常有用，為我們帶來了一些基本思想。 除了數字數據，文本數據還可以帶來一些有用的信息。 例如，哪種類型的廣告標題或主題更具吸引力，而消費者更有可能點擊它。

文字數據 (Text Data)

People prefer topics like team-oriented, fully configurable, and context-sensitive, etc. We also can see that some topics that consumers do not really interest in.

人們更喜歡面向團隊的主題，完全可配置的和上下文相關的主題等。我們還可以看到，一些消費者并不真正感興趣的主題。

相關性 (Correlations)

People aged around 40 who spend less than 80 minutes are more likely to click the ad. In Class 1, the mean age is around 40, and the daily internet usage range is between 100–200.

40歲左右，花費時間不到80分鐘的用戶更有可能點擊該廣告。 在第1類中，平均年齡約為40歲，每天的互聯網使用范圍為100-200。

造型 (Modeling)

Random Forest was applied in this project. After tuned hyperparameters with Grid Search and fitted the model, the accuracy and F1 score are up to 97%.

隨機森林應用于該項目。 在使用Grid Search調整超參數并擬合模型后，準確性和F1得分高達97％。

The feature importance chart shows that daily internet usage, daily time spent on the site, age, and area income play important roles for consumers' decision — “ Click it or not”!

功能重要性圖表顯示，每天的互聯網使用量，每天在網站上花費的時間，年齡和地區收入對消費者的決定起著重要的作用-“單擊或不單擊”！