Abstract
Every day we experience an unprecedented amount of data growth from numerous sources, which contribute to big data. As these data is growing rapidly in terms of volume, velocity, and variability, these impose great challenges on analytics framework and computational resources, making the overall analysis difficult to extract meaningful information in a timely manner. Thus, to harness such kind of challenges, developing an efficient big data analytics framework is a non-trivial research. Machine learning (ML) and deep learning (DL) algorithms are being in use in such an analytics framework to address these challenges by exploiting non-linear relationships from very high-dimensional datasets. Apache Spark has been in use as a popular big data processing arsenal, which helps develop iterative tasks of machine learning, having multi-algorithm open source, distributed ML library called Spark MLlib. Nevertheless, considering real-world research problems, long short-term memory (LSTM) is the effective approach to deal with the reduced accuracy problem due to vanishing and exploding gradient in conventional deep architectures. In this paper, we propose an efficient analytics framework, which technically a progressive machine learning frameworks, which is merged with multi-layer perceptron (MLP) and LSTM using two-stage cascade structure in order to enhance the classification accuracy. Our proposed architecture enables us to organize big data analytics in a scalable and efficient way. Experimental results outperform state-of-the-art approaches with a high-level of classification accuracy.