Laptop Price Range Prediction with Machine Learning Methods

Abstract

Prices forecasting, and price range estimation studies are very important for laptops, which have a very wide usage area, number of users and a large market share. Most existing price prediction studies use regression-based methods to estimate a concrete value for price. However, for many real-world applications, it is much more practical to predict a price class (or range). Although there are many studies on laptop price prediction in the literature, there is only one study on laptop price range prediction. The fact that the prices are divided into three different classes in this study does not overlap much with the laptop price range prediction problem in the real world. In addition, very few machine learning methods have been tested on the laptop price range prediction problem. To overcome these problems and contribute to the literature, a dataset previously used for laptop price prediction was adapted to be used for laptop price range prediction and the dataset was optimized for laptop price range prediction by applying preprocessing steps such as data cleaning, feature engineering and label encoding. Then, price range predictions were produced with machine learning methods such as random forest, histogram-based boosting, extra trees and catboost classifiers. When the success of the classifiers was tested, the best classifier was histogram-based boosting classifier with 70% accuracy.