Smartphone Price Prediction via Enhanced Interaction Features

This study fixes the problems of old feature methods for smartphone price prediction. It uses a two-step feature engineering method. This method mixes market knowledge with statistical checks. It builds interactive features from market logic. Then, it picks five key features using significance tests and overall evaluation. Tests with four models including linear regression, support vector regression, random forest, and XGBoost show the better feature set improves predictions for all models. The R² score goes up by 59.9% for linear regression and 4.9% for random forest. More tests show this method can improve results by up to 32.2% more than old feature selection ways. This work gives a useful feature engineering method for smartphone price prediction. It shows why feature quality matters and gives a guide for building features in tough market prediction jobs.