Class Imbalance: Why Stratify Matters in Machine Learning

Moving beyond model.fit()—building Gradient Descent from scratch. 🤖 I’ve been spending time lately digging into the mathematical foundations of Machine Learning. While libraries like Scikit-Learn make it easy to implement linear regression in two lines of code, I wanted to see if I could replicate those results by building a Gradient Descent algorithm from the ground up in Python. In this video, I’m: Defining the cost function (Mean Squared Error). Calculating partial derivatives to update weights ($m$) and bias ($b$). Fine-tuning the learning rate and iterations to reach global minima. Comparing my manual results against the LinearRegression class from Sklearn. The result? A near-perfect match! Understanding the "why" behind the "how" is making me a much better developer as I work on more complex computer vision projects. #MachineLearning #Python #DataScience #GradientDescent #AI #CodingLife

Can we predict customer subscription using Machine Learning? In this project, I built a Decision Tree Classifier to predict whether a customer will subscribe to a term deposit. 🤖 Model Used: Decision Tree (max_depth=5) 📊 Steps: • Removed data leakage • Encoded categorical features • Trained & evaluated model • Analyzed feature importance 📈 Result: Achieved strong classification performance and clear interpretability. 🛠 Tools: Python | Scikit-learn | Matplotlib 🔗 GitHub: https://lnkd.in/dtvPETN3 #MachineLearning #DecisionTree #DataScience #AI

🚀 Built a Machine Learning Model to Solve a Real Classification Problem Recently worked on an end-to-end ML project where I: • Cleaned and preprocessed raw data • Performed detailed exploratory data analysis • Engineered meaningful features • Trained and evaluated multiple classification models • Optimized performance using proper validation techniques What stood out most? Model performance improved significantly after proper feature engineering and handling class imbalance — not just from switching algorithms. This project reinforced something important: Good ML isn’t about trying every model. It’s about understanding the data first. Tech used: Python, Pandas, Scikit-learn, Matplotlib, SQL More projects coming soon 👀 #MachineLearning #DataAnalytics #Python #AI #LearningInPublic #WomenInTech

🔥 Day 11 — The 30-Day AI & Analytics Sprint with Instant Software Solutions A small line of Python… A big lesson about memory. matrix = [[1, 2, 3]] * 3 flat = [num for row in matrix for num in row] Looks clean, right? But here’s what’s really happening 👇 🧠 The Hidden Problem matrix = [[1, 2, 3]] * 3 This does NOT create three separate lists. Python creates one inner list, then repeats the reference to it three times. All rows point to the same memory address. So if you run: matrix[0][0] = 99 The entire matrix becomes: [[99, 2, 3], [99, 2, 3], [99, 2, 3]] Because you didn’t copy the list… You copied the reference. ✅ Why Flattening Still Works flat = [num for row in matrix for num in row] This nested list comprehension: • Loops through each row • Then loops through each number • Builds a brand new list Output: [1, 2, 3, 1, 2, 3, 1, 2, 3] It works because we’re reading values — not modifying them. 💡 The Correct Way If you want independent rows: matrix = [[1, 2, 3] for _ in range(3)] Now each row has its own memory space. 🎯 Lesson of the Day Multiplying nested lists duplicates references — not objects. Understanding this detail is what separates writing code that works… from writing code that scales safely. #100DaysOfCode #AISprint #MachineLearning #DataScience #Programming #SoftwareEngineering #BackendDevelopment #The_30_Day_AI_&_Analytics_Sprint #Python #AI #DataAnalysis #LearningJourney 🚀

Starting my journey into AI & Machine Learning I completed my first data analysis project using Python. In this project, I built a script that: ✅ Loads a CSV dataset ✅ Calculates Mean, Median, Mode and Standard Deviation ✅ Visualizes data distribution using a histogram This experience helped me understand an important lesson — before building Machine Learning models, understanding data statistically is essential. Tools & Technologies: • Python • Pandas • NumPy • Matplotlib • Git & GitHub Through this project, I learned how data analysis forms the foundation of AI systems. 🔗 Project available on GitHub: https://lnkd.in/g_-ZPRdb Next step is deeper exploration into data preprocessing and machine learning concepts. #Python #DataScience #MachineLearning #AI #LearningJourney #GitHub #BeginnerToEngineer

Understanding ColumnTransformer in Machine Learning When working with real-world datasets, we often have numerical + categorical features together. Applying the same preprocessing to all columns is not correct. That’s where ColumnTransformer from scikit-learn comes in! 🔹 It allows you to apply different transformations to different columns in a single pipeline. 🔹 It keeps preprocessing clean, organized, and production-ready. 🔹 It avoids data leakage when used with Pipeline. Example: Apply Standardization to numerical features Apply OneHotEncoding to categorical features Combine everything into one transformed dataset This makes your ML workflow: ✔️ Cleaner ✔️ More efficient ✔️ Scalable 💬 Question: Have you used ColumnTransformer in your ML projects? What challenges did you face? Github : https://lnkd.in/dee_ZATE #MachineLearning #DataScience #Python #ScikitLearn #FeatureEngineering

I’m excited to share my recent project I worked on with my partner Khadija Azam 🧭 “Visualizing Uninformed Search Algorithms: An AI Pathfinder” In this project, I implemented six classical uninformed search algorithms and visualized how each one explores a constrained grid environment in real time using Matplotlib. What started as a simple goal, moving from Start (S) to Target (T) while avoiding obstacles, turned into a fascinating study of how differently each algorithm “thinks.” Algorithms implemented: • Breadth-First Search (BFS) • Depth-First Search (DFS) • Uniform Cost Search (UCS) • Depth-Limited Search (DLS) • Iterative Deepening DFS (IDDFS) • Bidirectional BFS ✨ The visualization: Bidirectional BFS dramatically reduces search effort when start and goal are far apart BFS/UCS remain the safest choices for guaranteed shortest paths IDDFS provides an excellent memory-efficient alternative DFS behavior is highly sensitive to neighbor order The vertical wall constraint created interesting bottleneck Read the full Medium article here: 👉 Visualizing Uninformed Search Algorithms: An AI Pathfinder in Python https://lnkd.in/dD-XzwMh GitHub Link: https://lnkd.in/dfVyvp7c #ArtificialIntelligence #Python #Pathfinding #Algorithms #AIProjects #ComputerScience #Matplotlib #MachineLearning

🚀 Exploring Machine Learning classification with Decision Trees! In this quick walkthrough, I'm using Python and Scikit-learn to build and evaluate a DecisionTreeClassifier. It's always great to revisit the fundamentals and get hands-on with classic datasets like the Titanic survival data. 🚢 Here is a quick look at my workflow: 🧹 Data Preprocessing: Dropping unnecessary features, handling missing values, and converting categorical data into numerical data using LabelEncoder. ✂️ Data Splitting: Using train_test_split to ensure the model is evaluated on unseen data. 🌳 Model Training: Fitting the Decision Tree to the training set, checking the accuracy score, and making predictions! Building a strong foundation in these core ML concepts is key to tackling more complex AI challenges. What’s your go-to algorithm for classification tasks? Let me know in the comments! 👇 #MachineLearning #DataScience #Python #ScikitLearn #ArtificialIntelligence #DecisionTrees

We all use optimisers in Machine Learning — but how often do we actually see them working? I built Gradient Descent from scratch in Python, implementing: • Vanilla Gradient Descent • Momentum • Learning Rate Decay • RMSprop No ML libraries. Just NumPy, math, and curiosity. I visualised the entire training process — loss curves, weight & bias updates, parameter movement, and even full training animations. Watching the line slowly move toward the true parameters makes the theory feel real. Big takeaway? Optimisers aren’t magic. They’re disciplined update rules applied repeatedly. I did take GPT’s help in structuring parts of the code — AI speeds things up, but real understanding comes from building and experimenting yourself. Code here: https://lnkd.in/d4maNaR4 #MachineLearning #GradientDescent #Python #AI #LearningInPublic #DeepLearning #NeuralNetworks #ArtificialIntelligence #MLResearch #LearningDynamics #Optimization #GradientDescent

🚀 Comparative Model Evaluation on Wine Dataset | Machine Learning I recently performed a structured comparative analysis of multiple supervised learning algorithms on the Wine classification dataset using Python and Scikit-Learn. 📌 Objective Identify the best-performing classification model based on cross-validated accuracy. 🔬 Methodology Dataset: UCI Wine Dataset (multi-class classification) Evaluation Strategy: 10-Fold Cross Validation (KFold, shuffle=True, random_state=42) Metric: Accuracy Score Visualization: Boxplot comparison of cross-validation results 🤖 Models Compared Logistic Regression K-Nearest Neighbors (KNN) Decision Tree Support Vector Machine (SVM) Gaussian Naive Bayes 📊 Results SVM: ~98% accuracy Logistic Regression: ~97% accuracy KNN: ~96% accuracy Decision Tree: ~94% accuracy Naive Bayes: ~93% accuracy (Replace with your exact values.) 🏆 Best Model Support Vector Machine achieved the highest mean cross-validation accuracy, demonstrating strong generalization capability for this dataset. 💡 Key Learnings Importance of cross-validation over single train-test split Model variance analysis using boxplots Performance comparison beyond intuition Practical workflow for model selection This exercise strengthened my understanding of model evaluation pipelines and systematic algorithm benchmarking. Looking forward to applying similar comparative frameworks to larger real-world datasets. #MachineLearning #DataScience #Python #ScikitLearn #ModelSelection #CrossValidation #ArtificialIntelligence