New Edition of Regression Textbook Now Available

As someone in the people analytics space and interested in learning more about Bayesian and causal methods, the Bayesian and causal inference sections are exciting additions! Great to see these techniques being discussed more in people analytics. #peopleanalytics #datascience

COVARIANCE tells you the direction, but it's terrible at telling you the strength because it’s sensitive to scale. 🚀 Why Covariance Isn't Enough for Machine Learning I’ve been diving deep into the math behind ML, and today’s lesson was a "lightbulb moment" regarding Covariance vs. Correlation. 💡 In the code snippet below, I plotted two identical datasets, but I scaled the second one by a factor of 2. The Observation: Plot 1: Covariance = 1148.75 Plot 2: Covariance = 4595.0 Even though the "relationship" between the points is exactly the same (the shape of the scatter remains identical), the covariance quadrupled! The Takeaway: Covariance is scale-dependent. If you change your units from meters to centimeters, your covariance will explode, making it very difficult to compare relationships across different features or datasets. This is why we need Correlation (r): Correlation is just "standardized" covariance. It forces the value to stay between −1 and +1. Whether I measure in miles or millimeters, the correlation remains exactly the same. In Machine Learning, understanding this distinction is crucial for feature selection and understanding how variables truly interact without being misled by their magnitude. Math check: Corr(X,Y)=σX σY Cov(X,Y) By dividing by the standard deviations, we strip away the scale and keep only the relationship. 📉 #MachineLearning #DataScience #Mathematics #LearningJourney #Python #Statistics

🏠 California House Price Prediction using Machine Learning I recently worked on a California Housing Price Prediction project using Machine Learning. This project was built by following concepts from the book Hands-On Machine Learning with Scikit-Learn, Keras, and TensorFlow and implemented using Scikit-learn. What I learned and applied: 1. Data exploration and feature engineering. 2. Handling missing values and categorical data. 3. Building pipelines using Scikit-learn. 4. Training models like Linear Regression, Decision Trees, and Random Forest. 5. Evaluating models using cross-validation ✅ The project helped me clearly understand how machine learning models can predict real-world prices based on data such as location, population, and housing features. Grateful to the amazing learning resource and excited to apply these skills to more real-world problems!  #MachineLearning #ScikitLearn #DataScience #Python #CaliforniaHousing #LearningJourney #HandsOnML

Why many machine learning models fail in finance has nothing to do with algorithms. It starts with the data. I just published my first project in a personal portfolio on Financial Machine Learning, focused on one foundational topic: stylized facts of financial returns. Using real market data, I explore: Why financial returns are not normally distributed The presence of heavy tails and extreme events Volatility clustering and persistence Why standard ML assumptions often break down in financial time series Before applying complex models, it’s critical to understand the statistical structure of the data we’re modeling. Ignoring these properties leads to overfitting, fragile strategies, and misleading backtests. This project is inspired by Machine Learning and Data Science Blueprints for Finance and Advances in Financial Machine Learning. 📊 (attach the QQ-plot image here) I’m building this portfolio publicly as a way to deepen my understanding of financial ML and algorithmic trading, step by step. #FinancialMachineLearning #DataScience #QuantitativeFinance #AlgorithmicTrading #Python #MachineLearning

Stop filling missing data with just the Mean or Median. 🛑 That is "Univariate" thinking. It ignores the complex relationships between your features. To truly boost your model's accuracy, you need a Multivariate approach. 💡 Enter the KNN Imputer. Instead of blindly guessing, KNN acts like a detective. It scans all dimensions of your data to find the "nearest neighbors" and uses their patterns to intelligently fill the gaps. I broke down the entire process in this carousel: 🔹 The Logic: How spatial logic finds missing values. 🔹 The Math: Modified Euclidean distance for NaN values. 🔹 The Code: Implementation using sklearn. 🔹 The Trade-off: Why "Distance" weighting usually beats "Uniform." It is computationally heavier, but for small-to-medium datasets, the accuracy gain is worth it. 🔗 Get the code and resources here: https://lnkd.in/dbrEahc9 #DataScience #MachineLearning #Python #DataEngineering #AI #DataAnalysis #BigData #ScikitLearn #FeatureEngineering

Find min/max rows by another column in one line with Polars v1.37.0 🍰 Finding the row with the minimum or maximum value based on another column requires sorting, grouping, or complex filter expressions. Polars v1.37.0 adds min_by and max_by expression methods. These methods find minimum or maximum values based on any column in a single, readable expression. Get min_by and max_by by upgrading: "pip install -U polars". ☕️ Run this code: https://bit.ly/49EPy — 📬 This post is part of the CodeCut newsletter, which focuses on data science, AI, and open-source tools. Subscribe here → https://bit.ly/4p4WfZX #Python #DataEngineering #DataScience

📊 Exploring Multiple Linear Regression with Real-World Data 🏠📈 I recently worked with the California Housing dataset from scikit-learn to understand how multiple factors influence house prices. Using Multiple Linear Regression, I trained a model to predict housing values based on features like income, location, population, and housing characteristics 📍💰🏘️ The model achieved an R² score of ~0.59, which means it explains nearly 60% of the variation in house prices — a strong baseline for real-world data where many hidden factors exist 🌍📊 This helped me understand: 🔹 How multiple independent variables interact with a target 🔹 How to interpret R² and Adjusted R² 📌 GitHub 🔗: https://lnkd.in/gcNTNwFN #MachineLearning #DataScience #LinearRegression #AI #Python

💎📊 Predicting Diamond Prices with Machine Learning Built an end-to-end Diamond Price Prediction project using multiple regression models, focusing on understanding how different features influence pricing and how model choice impacts performance. I experimented with various regression techniques and compared their performance using proper evaluation metrics. Among all the models, Random Forest Regressor emerged as the best-performing model, achieving ~97% accuracy, highlighting the power of ensemble methods in capturing complex, non-linear relationships. 🔹 Project highlights: • Data preprocessing & feature understanding • Training and evaluating multiple regression models • Model comparison based on performance • Final model selection using Random Forest This project reinforced an important takeaway: 👉 Model selection and evaluation matter as much as model training. 🔗 GitHub Repository: https://lnkd.in/dZMMZT9x More ML projects coming as I continue building and refining my data science skillset 🚀 Feedback and suggestions are welcome! #MachineLearning #Regression #RandomForest #EnsembleLearning #Python #ScikitLearn #DataScienceProjects #LearningInPublic #AspiringDataScientist #MLProjects

My first hands-on step into Machine Learning. 📌 What I worked on: Predicting aqueous solubility (LogS) of molecules using molecular descriptors. 📌 What I did (from scratch): Loaded a real-world dataset using pandas Separated features and target variable (logS) Split data into training & test sets Trained: Linear Regression Random Forest Regressor Evaluated models using MSE and R² score Visualized Experimental vs Predicted LogS to understand model behavior 📊 The plot helped me clearly see how predictions align with actual values and where errors occur — a small but powerful insight for a beginner like me. 📌 What I learned: How model performance differs on training vs test data Why evaluation metrics matter more than just “running the code” Visualization makes model results much easier to interpret This isn’t a groundbreaking project — and that’s okay. It’s a learning milestone, and I’m sharing it to stay consistent and accountable while building my ML fundamentals. Next steps: feature importance, tuning models, and trying more datasets #MachineLearning #DataScience #LearningByDoing #Python #ScikitLearn #BeginnerJourney #MLBasics

Day 1: Explaining Machine Learning Algorithms (Simple Series) Ever wondered how machine learning predicts something like “exam score” from “study hours”? That’s exactly what Simple Linear Regression does: it fits a best-fit straight line to model the relationship between one input x and one output y. Core idea: y = β0 + β1x + ε Where: - y = predicted outcome (target) - x = input feature - β0 (intercept) = the baseline value of y when x = 0 - β1 (slope) = how much y changes when x increases by 1 - ε (error) = the part the line can’t explain (noise / randomness) How it learns (OLS): It chooses β0 and β1 by minimizing the squared gap between actual values and predicted values. Before trusting it, quick checks (assumptions): - Linearity (relationship is roughly a straight line) - Independent errors - Constant variance (homoscedasticity) - Errors are normally distributed (helps with inference) #MachineLearning #DataScience #ArtificialIntelligence #Python #Analytics