Mine & Rock Prediction by Ensemble Machine Learning Algorithms

Scientists and the military use submarines to go deep under the ocean water. Defense use submarines to guard the oceans. It is usually enormous and provide greater security. But marines have threats as obstacles and may be mistakenly recognized the rock as mine or mine as a rock. The SONAR technique has exploited the discovery of rocks and minerals. Objects can be detected using high-resolution images, but unique methods are required for identification. This study works on a dataset that differentiates the rocks and mines using SONAR signals. Under 60 various angles using SONAR signals, frequencies are recorded. Machine learning standalone classifiers, cross-validations, and bagging classifiers are applied on the data set to build the models. Standalone algorithms have challenges like overfitting, under fitting, feature selection, generalization on new data, etc. Bagging with cross-validation on standalone algorithms overcome overfitting, stability, bias reduction. The objective is to improve the model performance, enhance reliability and generalization ability of machine learning models. Random forest works best in all categories compared to other algorithms.