click here to download the base paper
click here to download the abstract project
ABSTARCT:
The paper is concerned with statistical models to forecast resale prices of used cars. An empirical study is performed to explore how different degrees of freedom in the modeling process contribute toward forecast accuracy. First, a comparative analysis of alternative prediction methods evidences that random forest regression is particularly effective in resale price forecasting. It is also shown that the use of linear regression, the prevailing method in previous work, should be avoided. Second, empirical results evidence the presence of heterogeneity in resale price forecasting and identify methods that can automatically overcome its detrimental effect on forecast accuracy. Finally, the study confirms that the sellers of used cars possess informational advantages over market research agencies, which enable them to forecast resale prices more accurately. This implies that sellers have an incentive to invest into an in-house forecasting solution, instead of basing pricing decisions on externally generated residual value estimates.
INTRODUCTION:
In many developed countries, it is common to lease a car rather than buying it outright. A lease is a binding contract between a buyer and a seller (or a third party – usually a bank, insurance firm or other financial institutions) in which the buyer must pay fixed installments for a pre-defined number of months/years to the seller/financier. After the lease period is over, the buyer has the possibility to buy the car at its residual value, i.e. its expected resale value. Thus, it is of commercial interest to seller/financiers to be able to predict the salvage value (residual value) of cars with accuracy. If the residual value is under-estimated by the seller/financier at the beginning, the installments will be higher for the clients who will certainly then opt for another seller/financier. If the residual value is over-estimated, the installments will be lower for the clients but then the seller/financier may have much difficulty at selling these high-priced used cars at this over-estimated residual value Predicting the resale value of a car is not a simple task. It is trite knowledge that the value of used cars depends on a number of factors. The most important ones are usually the age of the car, its make (and model), the origin of the car (the original country of the manufacturer), its mileage (the number of kilometers it has run) and its horsepower. Due to rising fuel prices, fuel economy is also of prime importance. Unfortunately, in practice, most people do not know exactly how much fuel their car consumes for each km driven. Other factors such as the type of fuel it uses, the interior style, the braking system, acceleration, the volume of its cylinders (measured in cc), safety index, its size, number of doors, paint colour, weight of the car, consumer reviews, prestigious awards won by the car manufacturer, its physical state, whether it is a sports car, whether it has cruise
control, whether it is automatic or manual transmission, whether it belonged to an individual or a company and other options such as air conditioner, sound system, power steering, cosmic wheels,
GPS navigator all may influence the price as well. Some special factors which buyers attach importance in Mauritius is the local of previous owners, whether the car had been involved in serious accidents and whether it is a lady-driven car. The look and feel of the car certainly contributes a lot to the price. As we can see, the price depends on a large number of factors. Unfortunately, information about all these factors are not always available and the buyer must make the decision to purchase at a certain price based on few factors only. In this work, we have considered only a small subset of the factors mentioned above. More details are provided in Section III. This paper is organised as follows. In the next section, a review of related work is provided. Section III describes the methodology while in section IV, we describe, evaluate and compare different machine learning techniques to predict the price of used cars. Finally, we end the paper with a conclusion with some pointers towards future work.
