The technology of data mining has been at the forefront of medical research. Not only has it achieved remarkable results in assessing patients' risks, but it has also been used in numerous models to make clinical decisions and predict diseases. Using various data mining models, raw data can be transformed into useful information that can be analyzed logically and scientifically to produce accurate predictions and decisions. The selection of features that are more important than others presents a challenge when it comes to disease prediction. To improve the prediction's accuracy, subset feature selection is carried out. The study compared various classification models to select the most important features. Correctness is checked by extracting, loading, transforming, and analyzing the data. Based on the results, the Multi-Layer-Perceptron (MLP) neural network, Support Vector Machine (SVM), random forest, Particle Swarm Optimization (PSO-SVM), and Bayesian network are compared to other data mining models. In order to increase models' correctness and accuracy, they are cross-validated using both 10 fold and 5 fold methods. Exactness is assessed at 86.26%, 66.06%, 75.15, 78.11% and 94.62 % for irregular woodland, Bayesian organization, SVM, MLP-brain organization and PSO-SVM. As indicated by referenced processed rules most elevated precision for anticipating liver cirrhosis sickness is anticipated by half and half PSO-SVM