Araç üzerinde yapılan modifikasyonların yakıt tüketimine olan etkisinin makine öğrenmesi modeli kullanılarak gözlemlenmesi için yeni bir yaklaşım

Vehicles have become indispensable in modern societies. Although they provide great convenience, they also come with a lot of serious problems. Especially, vehicles with high fuel consumption and low efficiency have negative impacts on the world and vehicle users. Vehicles exhaust harmful gases by burning fuel, which affects the environment by causing air pollution. High fuel consumption also affects the driver's financial situations in a negative manner. Vehicle producers continue to investigate, and research means for decreasing fuel consumption and increasing efficiency of internal combustion engines but observing effects of any recent modifications on a vehicle requires expensive and long-term test processes. The main aim of this study is decreasing the required duration to a minimum and providing an easy comparison with the state of the vehicle before modifications. In this thesis, we investigate various sensor data from vehicles and identified those, which have an effect on fuel consumption. The fuel consumption model of a vehicle is formulized via the data collected from vehicle sensors by taking environmental factors into consideration using various machine learning algorithms. Artificial Neural Network, Gradient Boosted Trees and XGBoost algorithms are utilized and their results are compared. According to the results, even the performances are observed to be close to each other, XGBoost algorithm provided the best results. After an appropriate model is obtained, impacts on the fuel consumption of modifications are observed compared to the previous state of the vehicle.