Scholarlink

Predicting the price of a car is always a challenging task for people as it depends on the features of the car. Features like number of cylinders, horsepower, city mileage per gallon, highway mileage per gallon, fuel type and brand of car play a crucial role in performance and price of a car. For a customer, it is always tough to apprehend the price of a car based on its features as a customer does not have sufficient knowledge of features of a car. Machine learning (ML) methods can help in this regard. In the present paper, we compare accuracies of four different ML methods: Linear Regression, Random Forest, Gradient Boosting, and XGBoost to predict the price of the car. We consider three different models having different feature sets and estimators. We also study the effect of parameter tuning by selecting the best hyper parameter values for each method using grid search. Automobile data set is used for the study which is openly available at Kaggle website. We found the best accuracy with Gradient Boosting irrespective of feature sets and estimators. We further observed improvement in accuracies using parameter tuning.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Comparing Car Price Prediction Accuracies of Different Supervised Machine Learning Methods with Parameter Tunings

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Raj Kishor Bisht,
  • Mayank Rawat,
  • Yogesh Joshi

摘要

Predicting the price of a car is always a challenging task for people as it depends on the features of the car. Features like number of cylinders, horsepower, city mileage per gallon, highway mileage per gallon, fuel type and brand of car play a crucial role in performance and price of a car. For a customer, it is always tough to apprehend the price of a car based on its features as a customer does not have sufficient knowledge of features of a car. Machine learning (ML) methods can help in this regard. In the present paper, we compare accuracies of four different ML methods: Linear Regression, Random Forest, Gradient Boosting, and XGBoost to predict the price of the car. We consider three different models having different feature sets and estimators. We also study the effect of parameter tuning by selecting the best hyper parameter values for each method using grid search. Automobile data set is used for the study which is openly available at Kaggle website. We found the best accuracy with Gradient Boosting irrespective of feature sets and estimators. We further observed improvement in accuracies using parameter tuning.