Liver Disease Prediction Using Machine Learning

Data Mining technologies have been widely used in the process of medical diagnosis and prognosis, extensively. These data mining techniques have been used to analyze a colossal amount of medical data. The steep increase in the rate of obesity and an unhealthy lifestyle eventually reflects the likelihood and the frequent occurrence of liver-related diseases in the mass. In this project, the patient data sets are analyzed for the predictability of the subject to have a liver disease based purely on a widely analyzed classification model. Since there are pre-existing processes to analyze the patient data and the classifier data, the more important facet here is to predict the same the conclusive result with a higher rate of accuracy. There are 5 distinct phases in this process. First, the min-max algorithm is applied to the original liver patient data set that could be collected from the UCI repository. In the second phase, significant attributes are demarcated by the use of PSO feature selection. This helps to bring out the subset of critical data, from the whole normalized datasets of liver patients. After this step, the third phase involves the usage of classification algorithms for comparative analysis and categorization. Accuracy Calculation is the fourth phase. It involves the usage of Root Mean Square value and a Root Error value. The fifth phase is the evaluation phase. Depending on the studies, a simple evaluation process is executed to preserve the integrity of a precise result reflection. J48 algorithm is considered to be a better performing algorithm when it comes to feature selection with an accuracy rate of 95.04%.