Analysis of Classification Algorithms for Oil Spill Recognition Using SAR Data

The research work examination extends the application of Synthetic aperture radar (SAR) data for recognition of oil let outs. The focus of this examination is on classing approaches for oil let outs detection using SAR imagery. Image processing techniques are used to derive feature vectors for ocean patch. The out-turn of the image processing is a fixed-length column vector for each dubious region used to classify whether the region contain an oil spill or not. The two classes of the data are oil spill and non-spill. The data of this study consists of standard imbalanced machine learning dataset. 937 cases with 48 derived attributes, a patch number, and a class description which has 896 cases as non-oil let out designate with class description 0 and oil spill with 41 cases is indicated by a class description. Basic EDA and data pre-processing are done before model building. Best classifier models are used for prediction of the data. The gradient boosting classifier providing the highest accuracy score of 98% and is selected as the best classifier for the oil prediction.