This project is based on a classification algorithm i.e. Naive Bayes which is run on a mobile dataset consisting of 2000 rows and 15 columns. It is a multi-class problem where mobile phones are classified in accordance with their price range. There are four classes of price ranging from 0 to 3, 0 indicating cheaper mobiles phones and 3 representing expensive mobile phones.
Univariate analysis is conducted to understand individual predictors and bivariate analysis is conducted to infer relationship between predictors with other predictors and target variable.
Important features are identified by Random Forest.