DESIGN INTELLIGENT MACHINE LEARNING CLASSIFIERS IN HEALTHCARE SECTORS FOR DIAGNOSIS ACCURACY

Abstract

Machine learning techniques are used widely in medical diagnosis to predict and classify tasks.
Machine learning techniques are used to diagnose diseases more accurately and more efficiently. The patient's
life-care machines and systems experience incremental improvements. This growth leads to an increase in the
average lifespan of humans. These health care systems are faced with many challenges, including misinforming
patients, privacy, inaccurate data, and lacks of medical information, classifiers to predict, and many other issues.
Various systems for diagnosing and predicting diseases have been created, including expert systems, clinical
prediction, decision support systems, and personal health records systems. This system is designed to assist
doctors in diagnosing diseases accurately. The diagnosis can be described as identifying the symptoms more
precisely. It is simple to treat the disease once the symptoms have been identified. These medical systems
require a lot of processing power and resources. Medical procedures are also computationally complex.
Researchers have two options when it comes to missing data in medical data: either they detect the problem and
remove the relevant data instances from the data set, or they use default methods like mean, median, neighbor,
etc., to fill the gap. Both ways do not produce optimal results. Outliers can also be found in the data, which
degrades the classifier's performance. Although it is not well explored, few researchers have focused on outlier
detection in medical data. This research also addresses two of the most well-known data problems, namely
missing value imputation and outlier. KMean++-based data imputation technology addresses the disappeared
value imputation problem. This technique validates data via clustering and also calculates missing data values.
A hybrid outlier detection method can detect the outlier. LS-SVM classifiers can determine the outcome. This
work provides a framework for diagnosing and predicting diabetes using LS-SVM classifiers.