Python Programming and Data Science projects

Compilation of Python programming and data analysis projects and challenges where I practice Python programming

Python Programming course at the University of Washington

This is a collection of Python programming and data analysis problems that I have practised in my Master of Science as part of the Python Programming course at the University of Washington: Python Programming at UW

The Data Science Course: Complete Data Science Bootcamp

This is a collection of data science and machine learning projects that I have practised as part of the Complete Data Science Bootcamp from 365 Careers on Udemy: The Data Science Course: Complete Data Science Bootcamp

1. Logistic Regression to predict Absenteeism at Work using Machine Learning:

Developed a logistic regression machine learning model using scikit-learn to predict excessive absenteeism at work based on health reasons, transportation expenses, family size, education level, and achieved a test accuracy of 74%. Conducted comprehensive data analysis, including creating dummies to categorize 28 absence reasons, extracting day and month from dates, and transforming numerical features with a custom scaler. Identified key predictors such as poisoning, diseases, transportation expenses, and household size. Created a module with class and methods to deploy the model on new observations to generate probability and predictions. Developed pattern visualizations in Tableau for age, absence reasons, transportation expenses and household size.

2. Log-linear Regression to predict Used Car prices using Machine Learning:

Developed a log-linear regression machine learning model using scikit-learn to predict used car prices based on specifications such as brand, mileage, engine volume, and year of production, achieving an Adj R squared of 75%. Conducted comprehensive data analysis, including descriptive statistics, creating dummies for categorical data such as brand, body, engine type, and registration, handling missing values, exploring probability distribution functions, detecting anomalies, handling outliers, and transforming inputs with StandardScaler. Also checked Ordinary Least Squares assumptions of linearity, no endogeneity, normality, homoscedasticity, no autocorrelation for continuous variables via matplotlib and scatter plots, and no multicollinearity by removing high-variance inflation factor variables. Developed scatter plots to visualize residuals between targets and predictions, created summary table for weights and bias, calculated percentage differences between targets and predictions and identified key predictors such as mileage and engine volume.