HYPERPARAMETER OPTIMIZATION OF DATA MINING ALGORITHMS ON CAR EVALUATION DATASET

Abstract

Data mining is the process of obtaining valuable data from large-scale data. Several algorithms are used for revealing the relationships between data and making accurate predictions. There are several cases that may affect the performance of these algorithms. One of these is choosing most suitable hyper parameters. To optimize these parameters provide us to improve algorithm results. In our project, we optimized hyper parameters of different data mining algorithms on car evaluation dataset for improving classification accuracy. Hyper parameter optimization was performed on support vector machines, k-nearest neighbor, random forest, multi-layer perceptron and gradient boosting algorithms. Results of hyper parameter values and default parameter values were compared. The calculations show that gradient boosting with hyper parameter optimization method produces best prediction of the car evaluation by 99.42%.