Hyper-Parameters Tuning for Hyperspectral Image Classification

Hyperspectral imaging is being applied in the field of Cosmology, farming, bioinformatics, medical devices, geology, physics, & monitoring are some of the fields covered. However, the hyperspectral data are multi-dimensional and noisy. Hence, it is challenging to classify data for meaningful use. Different machine learning algorithms are being used for classification. In this paper, three datasets namely such as Botswana, Pavia University Scene, Kennedy Space Center having spectral band of 145,103 and 176 respectively have been taken. Random forest and Support vector machines (SVM) algorithm have been used. The grid search method is used for tuning and finding out the best parameters for the respective model. The results obtained before and after tuning the hyperparameters have been compared. It has been discovered that the latter delivers superior results.