Input data characterization using machine learning and deep learning

Abstract

Seeing the current trends in Information technology, a large volume of heterogeneous data is produced widely across the world by means of social media sites such as Facebook, Instagram, Google plus, etc. and electronic gadgets used by humans for instance sensors. The generated data act as garbage and makes no sense until they are categorized. A strong need of data analytics or data filtering came into existence to filter the data into concrete categories. The generated data contain enormous features with huge dimensions. The high dimension data must be reduced to low dimension data to avoid curse of dimensionality and to build a better machine learning (ML) model. A ML model is built to perform the classification task that result into labeling of the data. In the first phase of work, a framework has proposed to process the data having features belonging to text and images using ML and DL algorithms. In the second phase of work, we have shown data related behavior using ML and DL framework. The proposed work is generic in sense of FS techniques and FE techniques using ML classifiers and DL classifiers. After carefully implementing the algorithms on the datasets we have evaluated the results taking accuracy as our performance metric. After carefully analyzing the accuracy it can be concluded that DL algorithms give better results than ML algorithms on both text and image dataset.