Pneumonia disease detection and classification system using naive Bayesian technique

Pneumonia is a chronic inflammation illness that affects both children and adults and is spread by various bacteria, viruses, and fungi. Since there are not  enough specialists and facilities to interpret the findings of lab-based diagnosis, resulting to several cases of Pneumonia-related deaths. When the  disease is discovered at an early stage as opposed to a later stage, it can be easily managed or controlled. The aim of the study is to create an effective  pneumonia disease detection and classification system that uses Naive Bayesian and random forest Algorithms. The hash-based function was applied to  train the model on X-ray chest samples from patients with pneumonia in order to improve detection accuracy and decrease classification errors. The  hashing-based function was employed to compute and convert X-ray image features to a corresponding numerical code or label stored in a relative  address and used as an array of reference given the associated values. The system was implemented using a future scaling technique that required the  use of a hash encoding algorithm for the categorical labels of the target variable, and it improved model performance. We validated and compared the  techniques in terms of accuracy and RMSE across different fine-tuned hyper-parameter values. The RF produced 97% with 3.33 error rate while NB  recorded 99.08% accuracy rate as the best with 0.020 RMSE value.