This project applies the ABC Analysis framework (80/20 Pareto principle) to an e-commerce dataset containing 112,650 transaction entries from a Brazilian e-commerce business spanning a four-year period.By categorizing products into A (High-Value), B (Mid-Value), and C (Low-Value) classes based on their cumulative revenue contribution, this study explores the direct relationship between an item's financial value and its purchase velocity. The analysis leverages dplyr for data segmentation and ggplot2 for descriptive visualizations.Additionally, the documentation highlights an essential data-cleaning and quality-assurance workflow—specifically addressing variable reference overrides and implementing a Logarithmic Scale ($Log_{10}$) transformation on boxplots to accurately resolve heavily skewed product lifecycles. The final visualizations provide clear, actionable insights into customer repeat-purchase behavior, offering distinct inventory management and capital optimization strategies for each product class.

This project delivers a comprehensive Customer Value Segmentation analysis for our online education platform, utilizing transaction data spanning 9,935 orders across a 12-month period (January 1, 2019 – December 31, 2019). By implementing a Recency, Frequency, and Monetary (RFM) framework, the analysis moves past aggregated metrics to map true student and institutional behavior. The objective is to transition from uniform marketing to high-precision, automated B2C workflows and high-touch B2B account management.

This project demonstrates a 'Global Model' approach to time-series forecasting using Python and Quarto. It features a robust pipeline for data preprocessing, multi-model benchmarking (Linear Regression, KNN, AdaBoost, SVR), and out-of-sample performance evaluation. The final implementation utilises Support Vector Regression (SVR) to handle non-linear demand spikes, achieving a stabilised forecast across a diverse dataset of over 3,400 observations.

This project demonstrates a 'Global Model' approach to time-series forecasting using Python and Quarto. It features a robust pipeline for data preprocessing, multi-model benchmarking (Linear Regression, KNN, AdaBoost, SVR), and out-of-sample performance evaluation. The final implementation utilises Support Vector Regression (SVR) to handle non-linear demand spikes, achieving a stabilised forecast across a diverse dataset of over 3,400 observations.