Deep learning methods are extensively applied to various fields of science and engineering such as speech recognition, image classifications, and learning methods in language processing. Similarly, traditional data processing techniques have several limitations of processing large amount of data. In addition, Big Data analytics requires new and sophisticated algorithms based on machine and deep learning techniques to process data in real-time with high accuracy and efficiency. However, recently, research incorporated various deep learning techniques with hybrid learning and training mechanisms of processing data with high speed. Most of these techniques are specific to scenarios and based on vector space thus, shows poor performance in generic scenarios and learning features in big data. In addition, one of the reason of such failure is high involvement of humans to design sophisticated and optimized algorithms based on machine and deep learning techniques. In this article, we bring forward an approach of comparing various deep learning techniques for processing huge amount of data with different number of neurons and hidden layers. The comparative study shows that deep learning techniques can be built by introducing a number of methods in combination with supervised and unsupervised training techniques.