Underwater mines are widely used by the marines or navy for the defense purposes or warfare. It gives great security, but the mines can be easily mistaken for rocks, this can be a great risk. So, for this purpose we need an accurate system to predict weather the object is a threat or not. To have a good accuracy, we need to have accurate data and good system to generate accurate results. This data will be used to train the system. This paper presents a method for the prediction of mines and rocks using sonar data. We used the data that is generated using sonar signals to record the various frequencies of underwater objects at 60 different angles. Then this data is used for training the system. The prediction model defined by logistic regression is used to predict marine objects as mines or rocks. Our objective here is to detect 100% of the mines while minimizing the incorrect mine classification.