An approach towards realizing liver cancer diagnosis using supervised machine learning techniques

Abstract

Now-a-days liver cancer is one of the most prevalent diseases that may be extremely fatal if not properly diagnosed at early stage. The research work has a notable social impact as it facilitates liver cancer diagnosis on the basis of statistical approaches and experimental performance of machine learning classifiers on ILPD(Indian Liver Patient Dataset) and BUPA liver datasets. The work embodies certain discovered facts. The liver cancer diagnosis can be governed by concept learning, artificial neural modeling, geomatric distribution and Cobb-Douglas model. The augmentation or expansion of features indicating liver cancer growth can be quantified and realized based on Markov property based state transition. Liver cancer detection can also be analyzed based upon the fundamental principle of information gain. The realibility and mean time to failure of liver cancer testing system can be carried out in the light of parallel system configuration. The factor leading to liver cancer can be sensed on the basis of weighted majority algorithms.The present objective is also to propose a method using supervised machine learning that can help the physician for accurate diagnosis of liver cancer. For experimental analysis two liver cancer datasets are used. The six diverse classifiers in machine learning are used for the experimental propose, i.e., Tree, Meta, Rules, Lazy, Function and Bayes. All the six classifiers are applied on the available datasets. The performance of the machine learning model is best described in terms of their accuracy, precision and recall. The highest accuracy is achieved by using rule classifier on the BUPA liver cancer data set and Lazy classifier on ILPD. newline