Biomedical Named Entity Recognition (B-NER) is the process of extracting medical entities such as, disease name, genes, proteins, treatment names, drugs, etc. This paper focuses on extracting disease and treatment entities. In the recent years, several methods have been proposed for the extraction of entities related to the biomedical domain. This paper puts forth a B-NER system that extracts disease and treatment entities from an imbalanced class label distribution using novel feature extraction techniques. The novel feature extraction techniques adhere some hierarchy that helps to achieve good results to extract the named entities. The proposed system uses two classifiers namely, Multinomial Naive Bayes (MNB) and Maximum Entropy (MaxEnt) model. The test dataset consists of 12904 entities out of which 1711 are biomedical entities which in turn contains 960-disease entities and 751-treatment entities. The proposed system achieves 78% in disease entities and 69% in treatment entities with novel feature extraction techniques.