Data Science Project: Predicting Competitive Auctions on eBay (CS 701)

This project analyzes a dataset of 1972 eBay auctions to predict whether an auction will be competitive, defined as having at least two bids. The solution includes data preprocessing steps such as creating dummy variables for categorical features like category, currency, end day, and duration. The dataset is split into training and validation sets to build and evaluate a predictive model. The analysis involves descriptive statistics, normality tests, and correlation analysis to understand the data. The project uses logistic regression to model the probability of an auction being competitive, incorporating variables such as seller rating, close price, open price, and the categorical variables. The solution includes the iteration history, omnibus tests, model summary, Hosmer and Lemeshow test, and contingency tables to assess model performance and identify key predictors of auction competitiveness. The project aims to distinguish competitive auctions from non-competitive ones using a data-driven approach.