Regression Algorithms for Intelligent Estimation of the Selling Price of Used Vehicles – A Machine Learning Approach

Transportation is undergoing a major change in this rapidly growing world. Technological progress has made automobiles a possibility making this small world even smaller. This means progress and advancements are leading to many people switching vehicles frequently in anticipation of better features. Sadly, in many countries, they are left in the dark about the best price their vehicle can fetch in the market and therefore are forced to settle for a price decided by the used-car outlets or their best instincts. This means that the owners of the vehicles may not always end up getting the best deal for their cars while selling. A second drawback to the existing approach, the selling price of their cars may not always be consistent with different dealers as they may offer different prices for the same car. A third drawback is the current scenario is not user friendly as the car owners have to travel from dealer to dealer to get to know the selling price of the car. This opens up the possibility of having a machine learning approach to this task of estimating the selling price of the pre-owned cars. Fortunately the availability of datasets with car features and their actual selling price of historical sales data makes it possible to develop a regression model using different techniques. In this paper, we therefore investigate the application of different regression algorithms to this task of estimating the best selling price of the car and use different metrics or cost functions such as Mean Square Error (MSE), Mean Absolute Error (MAE) etc. to evaluate the models’ performance to this task thereby helping arrive at the best algorithm for this task. It is our hope that the implementation of this algorithm can improve the scenario of car price estimation that is prevalent today.