<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by wafa Mribah on Medium]]></title>
        <description><![CDATA[Stories by wafa Mribah on Medium]]></description>
        <link>https://medium.com/@wafa.meribah?source=rss-180710256cd6------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*srQvj2enGv1S5WkjRLyg-w.jpeg</url>
            <title>Stories by wafa Mribah on Medium</title>
            <link>https://medium.com/@wafa.meribah?source=rss-180710256cd6------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Thu, 19 Nov 2026 14:35:32 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@wafa.meribah/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Embrace a Data-Driven mindset]]></title>
            <link>https://medium.com/@wafa.meribah/embrace-a-data-driven-mindset-f290fcdc0a7c?source=rss-180710256cd6------2</link>
            <guid isPermaLink="false">https://medium.com/p/f290fcdc0a7c</guid>
            <dc:creator><![CDATA[wafa Mribah]]></dc:creator>
            <pubDate>Mon, 27 Mar 2023 06:07:01 GMT</pubDate>
            <atom:updated>2023-03-27T06:07:01.688Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*EsJMKrmm_Ml71ZFU.jpg" /></figure><p><strong>In this article</strong></p><ul><li>You will find the top 3 ingredients for an AI-powered solution.</li><li><strong>Outcome:</strong> Why and how to embrace a data-driven solution?</li><li><strong>Business use-case: </strong>Fraud detection in E-commerce.</li><li><strong>Target audience:</strong> Beginners, Business leaders.</li><li><strong>Main character:</strong> You</li><li><strong>Supporting characters: </strong>Pam, Jim, Dwight and Michael</li></ul><p>You started a new job as the head of risk management at one of the biggest online retail companies in the Netherlands “<strong>Pol.com”! </strong>( No pun-intended)</p><p>Congrats!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/206/1*fdwX3D6QosiKmKe5eg4gXw.png" /><figcaption>Logo of Pol.com and Reasons why I am *not* a UI designer</figcaption></figure><p>Your division’s main responsibility is to do a case-by-case review of the orders that are considered a potential fraud. <br>Of all the orders, only less than 5% were actually reviewed by an expert! Red flag 🚩 ! Experts are often busy, have a high workload and this task is time consuming and repetitive. As a consequence, orders are either falsely canceled or falsely accepted and this is causing a significant financial loss for the company and it is threatening the consumer trust.</p><blockquote>“ Triggered by the growing use of digital wallets, the total cost of e-commerce fraud to merchants is likely to exceed $48 billion globally in 2023, from just over $41 billion in 2022 “ According to Business-Standards</blockquote><p>At this point you are probably thinking “why did I even take this job?🤔 ”. Fair, but somehow you rolled up your sleeves and you set your first departmental goal at the company: <strong>“increase work efficiency through improving the fraud detection process at Pol.com by 60%”</strong></p><blockquote>1st ingredient for AI-powered solutions: framing the problem and setting clear objectives.</blockquote><p>So now that you recognized the problem and defined how success looks like, you decided to organize a brainstorming session which went something like this:</p><ul><li>Pam: “ We should probably hire more people!”</li><li>Jim: “We can probably cancel all potential fraudulent orders and if they order again then they might be legit, just saying :( ”</li><li>Dwight: “Some historical data can help us determine whether the transaction will be fraudulent or not”</li><li>Michael: “Well, I am not sure why I am here…”</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/479/1*WAUKE2vCCgFeX_lQWSUQig.png" /></figure><p>This getting even more confusing so let ask an old Pal:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*DJaVG_XnsJsaUAsMEPU0cg.png" /></figure><p>ChatGPT has a point here! if we take a look at the bigger picture, you are trying to solve a task that requires lots of repeated decisions, and what do we do with repeated decisions?</p><p>We automate!</p><blockquote>2nd ingredient for an AI-powered solutions: Automation</blockquote><p>With that in mind, you started investigating the data-driven approach and worked along side the Data science department <em>(and of course Dwight) </em>to automatically detect fraudulent orders.</p><p>The DS team explained the process for creating a new AI-Powered solution so they sent over the following diagram:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*eJ5hHSUZNK4f1dl5Y6i3ug.png" /><figcaption>Standard Data Science Life cycle</figcaption></figure><p>The following data science approach consists of five main steps:</p><ol><li><strong>Data collection: </strong>This step is critical to this approach. The quality and quantity of the data will highly impact the outcome of the algorithm.</li></ol><blockquote>3rd ingredient for AI-powered solutions: Good data</blockquote><p>The team at Pol.com found several historical files which are useful for this use-case. The data contains information about: <br>- The customer: ID, email, history of their transactions.<br>- The order: order ID, order amount.<br>- Some other useful information: Device(Laptop or phone), Credit card(Visa or MasterCard) <br>- and of course,<strong> the label</strong> which tells us whether that historical order was a fraud or not ( translated as 1 for fraud and 0 for not)</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/836/1*1sKOzapwE-Yilzk1niCCew.png" /></figure><p>As you can see above, the data is messy and there are some missing values and irrelevant information that is why cleaning the data is a necessary task before going further into the process.</p><p>It is time for the Data science team to get their hands dirty and bring this idea to life. The team is using <strong>Jupyter notebooks </strong>to develop the solution.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/100/1*MZFMjbcTJ_1cKqqOK_qQAA.png" /></figure><p><strong>Why Jupyter? </strong>because it is simply a great open source tool that is highly used in the DS community. It works both as a standalone and integrated into famous DS platforms like: Amazon SageMaker, Google Collab and Kaggle kernels. It is a great environment to develop code and present results.</p><p>The team gave you a high-level explanation about the flow and how to work on Jupyter Notebooks</p><ol><li>Import the required libraries</li><li>Read and understand the data</li><li>Data clean-up and preparation</li><li>Exploratory Data Analysis</li><li>Model building</li><li>Model configuration ( or also called Hyperparameters Tuning)</li></ol><p><strong>Putting the solution in use</strong></p><p>Everything came together nicely. The DS team built and evaluated the solution, they presented it to you and other business leaders. The results and accuracy were impressive and you cannot wait to try it in real time.</p><p><strong>So how to actually use the solution in practice?</strong></p><p>The team drafted a simplified flow on how to use the solution they built and they let the door open for the business leaders to pick between a fully automated decision making that doesn’t require a human intervention or a semi-automated one that requires partial involvement from the Subject matter expert.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*AvcXqeKCCQ0slGubLAZSig.png" /></figure><h4>Wrapping-up</h4><p>In this article, I chose a realistic but challenging use case and explained how to think about the solution while embracing a data-driven mindset.</p><blockquote>The data science process is a systematic approach to solving a data problem. It provides a structured framework for articulating your problem as a question, deciding how to solve it, and then presenting the solution to stakeholders.</blockquote><ul><li>We framed the problem together and defined clear objectives from the very start.</li><li>We brainstormed few ideas and explained the reasons and ingredients behind an AI-powered approach.</li><li>We learned the Data science Lifecyle and applied it to our use case.</li><li>We talked about the steps of using a popular tool like Jupyter Notebook.</li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f290fcdc0a7c" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>