Comparative Analysis for Prediction of Pneumonia using Deep learning Methods

Pneumonia is basically an infection which can infect one or both the lungs of a person through their air bladders. Their sacs of such a person might get filled with pus, which in turn cause cough along with problems related to breathing and fever. Pneumonia is usually originated from various organisms such as viruses, fungi and bacteria. Pneumonia may be mild or even life-threatening in some situations. It usually turns out to be very serious for newborns and very young children, also for senior citizens having age more than 65 years, especially people alreadyhavingsomehealthissuesorenfeebleimmunesystems.Thisresearch focuses on comparing the best ways of using Machine Learning and Deep Learning for detecting Pneumonia using its different symptoms as features. For the purpose of this research, the data setthath as been use dcan be extracted from Kaggle website. It is a comparative study to compare which as pects of the disease should be considered for the best model. We compared various deep learning and machine learning models such as Random Forest and numerous Convolutional Neural Network architectures(VGG-16,InceptionV3,2:1 Architecture without using Batch Normalization andDropout,4:2ArchitectureusingBatchNormalizationandDropout,5Convolutional Blocks CNN with Batch Normalization and Max-pooling) for each and every feasible symptom to provide a holistic way of determining whether or not patients offers from Pneumonia. Key Word: Pneumonia,ConvolutionalNeuralNetwork,VGG-16,InceptionV3, Random Forest