<?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 Abdullah Zaki on Medium]]></title>
        <description><![CDATA[Stories by Abdullah Zaki on Medium]]></description>
        <link>https://medium.com/@abdullah9zaki?source=rss-c2cc3d980223------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*vYoPN0hYmLpRjnwl</url>
            <title>Stories by Abdullah Zaki on Medium</title>
            <link>https://medium.com/@abdullah9zaki?source=rss-c2cc3d980223------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Mon, 16 Nov 2026 21:45:04 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@abdullah9zaki/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[Next Best Advertisement Campaign Selector]]></title>
            <link>https://medium.com/@abdullah9zaki/next-best-advertisement-campaign-selector-62e2c3ea8180?source=rss-c2cc3d980223------2</link>
            <guid isPermaLink="false">https://medium.com/p/62e2c3ea8180</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[algorithms]]></category>
            <dc:creator><![CDATA[Abdullah Zaki]]></dc:creator>
            <pubDate>Mon, 03 Mar 2025 23:50:17 GMT</pubDate>
            <atom:updated>2025-03-03T23:50:17.697Z</atom:updated>
            <content:encoded><![CDATA[<p>Goal of the Algorithm</p><p>The goal of this algorithm is to determine the most effective advertisement campaign for my customers by analyzing their data and predicting which campaign type will lead to the highest conversions. Businesses use different campaign types — like social media ads, email marketing, or display ads — to engage customers, but not every type works equally well for everyone. By using historical data about my customers (like their age, location, and past purchases) and how they responded to previous campaigns, this algorithm predicts which campaign will perform best for new customers. It can either personalize campaigns for individual customers or choose one campaign for the whole group, depending on my strategy. The aim is to maximize conversions (e.g., purchases) while being flexible enough to consider costs or variety, helping me make smarter marketing decisions.</p><p>Step-by-Step Guide to Creating the Algorithm</p><p>Step 1: Collect the Necessary Data</p><p>•	What I Did: I gathered data from my past advertisement campaigns to see what worked before.</p><p>•	Details Included:</p><p>•	Customer ID: A unique number for each customer so I can track them.</p><p>•	Customer Attributes: Things like age, gender, location, and purchase history that describe my customers.</p><p>•	Campaign Type: The kind of campaign I used, like social media ads, email marketing, or display ads.</p><p>•	Conversion Outcome: A yes/no marker (1 for yes, 0 for no) showing if the customer bought something after seeing the ad.</p><p>•	Why: This data is the starting point — it shows me patterns of what campaigns succeed with which customers.</p><p>Step 2: Preprocess the Data</p><p>•	What I Did: I cleaned up the data to make it usable for the algorithm.</p><p>•	How:</p><p>•	Fixed missing info (e.g., used average age if missing) and removed weird outliers (e.g., an age of 999).</p><p>•	Scaled numbers like age or spending to a 0-to-1 range so they’re comparable.</p><p>•	Changed text like “social media” into numbers (e.g., 1 or 0) using one-hot encoding.</p><p>•	Why: This makes the data neat and ready for a computer model to understand.</p><p>Step 3: Choose a Machine Learning Model</p><p>•	What I Did: I picked a model to predict if customers will convert.</p><p>•	Options:</p><p>•	Logistic Regression: Simple and good for basic patterns.</p><p>•	Random Forest: Better for tricky or curvy patterns.</p><p>•	Neural Network: Great for big data with complex trends.</p><p>•	My Choice: I started with logistic regression because my data isn’t huge or super complicated yet.</p><p>•	Why: The model figures out how likely each campaign is to work for a customer.</p><p>Step 4: Train the Model</p><p>•	What I Did: I taught the model using my cleaned-up data.</p><p>•	How: I gave it customer attributes and campaign types as inputs, and conversion outcomes as the answer it should learn to predict.</p><p>•	Process: The model studied the data to spot which customer types convert for each campaign.</p><p>•	Why: This training lets the model make smart guesses for new customers.</p><p>Step 5: Make Predictions for New Customers</p><p>•	What I Did: I used the model to guess how new customers will respond.</p><p>•	How: For each customer, I plugged in their attributes and each campaign type to get a conversion probability (e.g., 20% for social media, 35% for email).</p><p>•	Why: These numbers show me which campaign might work best for each person.</p><p>Step 6: Select the Best Campaign Type</p><p>•	What I Did: I decided which campaign to use based on the predictions.</p><p>•	Options:</p><p>•	For Each Customer: Pick the campaign with the highest probability for that person (good for personal ads).</p><p>•	For All Customers: Average the probabilities across everyone for each campaign and pick the highest (good for one big campaign).</p><p>•	Why: This step picks the winner, matching my goal of targeting individuals or the whole group.</p><p>Step 7: (Optional) Consider Additional Factors</p><p>•	What I Did: I added extra checks to make the algorithm smarter.</p><p>•	How:</p><p>•	Costs: If I know campaign costs, I divided total predicted conversions by cost to find the best bang for my buck.</p><p>•	Variety: I added a small penalty for reusing the same campaign too much to mix things up.</p><p>•	Why: This keeps my budget in check and stops customers from getting bored with the same ads.</p><p>Step 8: Implement and Monitor the Algorithm</p><p>•	What I Did: I put the algorithm to work and kept an eye on it.</p><p>•	How:</p><p>•	Used it to pick my next campaigns.</p><p>•	Watched real results to see if it’s right.</p><p>•	Planned to retrain it later with new data as trends change.</p><p>•	Why: This makes sure it stays useful and accurate over time.</p><p>Example Scenario</p><p>Let’s say I have three campaign types: social media ads, email marketing, and display ads. I ran the algorithm on 100 new customers, and it predicted these average conversion probabilities:</p><p>•	Social media ads: 12%</p><p>•	Email marketing: 18%</p><p>•	Display ads: 9%</p><p>Result: The algorithm picked email marketing for the group because 18% is the highest average. If I’d done it for individuals, it might’ve picked different campaigns per person, but for one campaign, email wins.</p><p>Why This Works</p><p>•	Data-Driven: It uses my real customer data and past results, so it’s based on facts, not guesses.</p><p>•	Flexible: I can use it for personal campaigns or one big one, whatever I need.</p><p>•	Scalable: It’ll handle more data or fancier models as my business grows.</p><p>•	Customizable: I can tweak it for costs or variety, making it fit my goals perfectly.</p><p>By building this algorithm, I’ve made a tool that picks the best advertisement campaigns for my customers, boosting my results and saving time. It’s like having a marketing assistant that learns from my data</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=62e2c3ea8180" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>