This document summarizes a project report on using regression analysis to predict daily maximum carbon monoxide (CO) levels using air pollution sensor data from Kanpur, India between 2004-2005. The analysis developed a multiple regression model to forecast CO levels for the next week using variables like previous day's benzene, temperature, absolute humidity levels, and dummy variables for months and days of the week. The model explained 65% of variability in CO levels. Residual plots showed the model fit the data properly.