The blue revolution of the blue economy on the way to build a golden Bangladesh is now the demand of the time. The blue economy is sea-based. The economy of exploiting the vast resources of the oceans and their bottoms. This means that whatever is extracted from the sea if it is added to the country’s economy, will fall into the category of the blue economy. But the amount of resources of Bangladesh at the Bay of Bengal (BoB) has not yet been surveyed properly. If the number of marine cetaceans can be known, then proper steps of marine management can be taken to protect marine mammals. This paper deals with detecting marine cetaceans based on various machine learning classification algorithms such as Support Vector Machine (SVM), Decision Trees (DT) classifier, k-Nearest Neighbors (kNN) classifier, Artificial Neural Network (ANN) classifier, and Convolutional Neural Network (CNN) around the Swatch of No Ground (SoNG) in the BoB. At first, the possible marine cetaceans living around the BoB has been listed for the training purpose of classification algorithms. Then the dataset (both training and validation or test) being trained to classification algorithms have been created by extracting spectrogram images of the clicks, whistles or songs of listed marine cetaceans around the SoNG. Three types of test data such as original test data (OTD), synthetic test data (STD) and practical test data (PTD) have considered validating the proposed method. The test data retrieved from the original dataset is the OTD. The STD and PTD have been derived from the OTD. Then these algorithms have been trained with the training sets selected from created dataset for the detection and classification of marine cetaceans. After completing the training process, the proposed algorithm has been evaluated with three types of test data and recorded the output to analyze the performance in detection and classification of marine cetaceans. The detection process will be very challenging as there is a lot of noise in the sea. That’s why we tested our model by generating synthetic and practical clicks, whistles or songs of marine cetaceans and comparatively satisfactory results have been obtained for CNN algorithm. This algorithm has been successfully detected and classified the species of marine cetaceans with the accuracy of 96.60% for OTD (Recall=0.94, F1-score=0.93), 93.38% for STD (Recall=0.91, F1-score=0.90) and 90.79% for PTD(Recall=0.91, F1-score=0.90).