SBM for mobile phone price classification

Feature selection has been playing a significant role in many machine learning applications because of its omnipresent nature in any problem related to ML & DL. Extensive research is recorded in selecting best features. In this paper eight classifiers including SVC, KNN, ANN, RF, Decision Tree, Naive Bayes, Linear Discriminant Analysis, Logistic Regression models are applied with different feature selection algorithms including Pearson correlation, Chi-squared, Recursive feature elimination, Lasso, Tree-based, ANOVA, Mutual information, Feature importance to predict price category of mobile phones. In the proposed work after applying appropriate feature engineering techniques first we dropped the constant features using variance threshold. After that highly correlated features are dropped using Pearson correlation in order to get best features to train the model. From the results, it is evident that ANOVA-F test is working satisfactorily for the Kaggle dataset with best features. Among all the classifiers SVC with SBM classification accuracy outperforms other classifiers. When SBM is applied to SVC the accuracy is increased from 96.7% to 97.8%.