Estimating and predicting orbital decay of low Earth orbiting satellites is a challenging task. Orbital decay occurs due to various factors such as solar flux, geomagnetic flux variations, atmospheric drag, spacecraft height, spacecraft mass, the material of the spacecraft, its size and shape, spacecraft attitude, and spacecraft altitude. Some of these parameters are known, while others are not accurately known or cannot be predicted with high precision. Thus, analytically accurate decay prediction is always challenging. On the other hand, accurate orbital decay prediction is crucial to keeping the spacecraft close to a reference orbit so that it remains within its defined ground track. This paper analyzes the problem and explores the possibility of estimating orbital decay using machine learning algorithms to achieve better results. To predict decay more accurately, the onboard GPS receiver data has been taken. This data are highly accurate as they are derived after receiving at least four or more numbers of signals from Global Positioning System (GPS) satellites. This satellite’s orbital data, along with associated variables are used to train the machine for future position predictions. Many deep learning algorithms like Long Short-Term Memory (LSTM), Dense learning Network are considered based on suitability and implemented. This paper deals with actual satellite data, and the results are presented have been compared with the actual results. The results are discussed in the conclusion section and the best algorithm is chosen for implementation.