A novel fuzzy three-valued logic computational framework in machine learning for medicine dataset

For consideration of uncertainties of a medicine dataset, a new conceptual architecture fuzzy three-valued logic is introduced in this research work. The proposed concept is applied to the heart disease dataset for the assessment of heart disease risk in individuals. By comparison of three binary (0,1) input variables, the variables' uncertainties and their collective impact can be analyzed that provide complete information leading to better outcome prediction. The availability of a wide range of values ultimately modified the output binary variable thus now providing three values (0,0.5,1) instead of binary (0,1) values. The three types of output values are heart disease risk is absent (0), may be present (0.5), and present (1). The inclusion of an additional class which is heart disease risk may be present (0.5) can alert an individual to include healthy lifestyle factors to minimize any chances of the presence of risk of heart disease development due to the behavioral risk factors at least thus helping an individual for better decision making. Initially, a subset of artificial intelligence (AI) i.e. traditional machine learning techniques have been applied to the considered dataset. Subsequently, it is applied to the fuzzy three-valued modified considered dataset hence leading to the development of a hybrid fuzzy three-valued modified machine learning model. This integration increases the accuracy of machine learning techniques from 70 % to 99 %. Likewise, the computation time has also been optimized and provides results in less than 11s. Statistical analysis using the Wilcoxon signed rank test and validation using an application of the proposed method on datasets of different domains also depicts that the proposed methodology is better and computationally efficient. Moreover, it also provides sufficient information to the individual thus helping in better decision making to live a disease free life.