An Efficient Approach For Pathogens Detection in Bacterial Pneumonia

Abstract

Pathogens detection in Bacterial Pneumonia can be understood as a classification of types of Bacterial Pneumonia. The motive is to provide the pathogen-directed antibiotics at the right time through the automation of detecting the pathogens in Bacterial Pneumonia with Artificial Intelligence. This study has combined the pre-trained model with machine learning classifiers for binary classification. The pre-trained model VGG16 is fine-tuned and then the features are extracted from this model. For the classification purpose, three machine learning classifiers (SVM, Random Forest, and Bagging) are used to evaluate the performance of the model on two metrics Testing Accuracy and Validation Accuracy. To generalize the model we have also used one of the most widely used K-Fold cross-validation techniques. A comparison is also made between the existing methodology and ours on the same dataset used in this study. Our methodology performs better by achieving the highest testing accuracy of 90.8% with Bagging as a Classifier. VGG16 achieved the lowest testing accuracy of 82.63% whereas VGG16 with Support Vector Machine achieved 89.65% of testing accuracy and with Random Forest 88.77%.