A Rule-Based Deep Learning Method for Predicting Price of Used Cars

The growth of internet used automobile shopping has been tremendous in recent years. It is important for sellers to correctly evaluate the price of the used cars. From the literature, classification is an approach that assists in data interpretation by analyzing data and providing a second opinion on data among other tasks most of the classifiers used for car price prediction suffered from lower accuracy ratios and higher error ratios. In this work, a hybrid feature selection model for predicting the price of used cars was developed. The developed model includes a CfsSubsetEval to determine the best feature subset, rule-based GA to search for the best solution, and a neural network for the final price prediction based on the selected features. The developed model was implemented in a Java programming language. The evaluation showed that the developed hybrid feature selection model achieved better performances with the correlation coefficient, mean absolute error and root mean square error of 0.9818, 1.1809, and 1.5643 respectively, compared to other related state-of-the-art algorithms. The results validated the efficiency of the developed model for used car price prediction.