Deep learning is a power full method for the analysis of the dataset. It can be applied in variety of the problems like regression, classification and clustering. Regression and classification comes under supervised learning techniques. Most of the real world data and problems come under supervised category. In recent years, there has been a lot of research on the categorization of datasets, and numerous strategies have been put out to solve the classification issue. But the majority of these techniques don’t extract deep characteristics in a hierarchical way. The authors for the first time apply deep learning to data classification in this paper. They initially assess the efficiency of employing stacked autoencoders before suggesting a novel strategy that takes into account geographical information. In order to attain the best classification accuracy, they also suggest a novel framework that integrates deep learning, principal component analysis, and logistic regression. It has been shown that this framework offers competitive performance on frequently used data.