Indonesian Journal of Electrical Engineering and Computer Science Vol. 24, No. 1, October 2021, pp. 564~569 ISSN: 2502-4752, DOI: 10.11591/ijeecs.v24.i1.ppa564-569  564 Journal homepage: http://ijeecs.iaescore.com Machine learning based outlier detection for medical data R. Vijaya Kumar Reddy 1 , Shaik Subhani 2 , B. Srinivasa Rao 3 , N. Lakshmipathi Anantha 4 1,3 Department of IT, Lakireddy Bali Reddy College of Engineering (Autonomus), Mylavaram, India 2 Department of IT, Sreenidhi Institute of Science and Technology (Autonomus), Telangana, India 4 Department of CSE, Malla Reddy Engineering College (Autonomus), Telangana, India Article Info ABSTRACT Article history: Received Jan 10, 2021 Revised Aug 8, 2021 Accepted Aug 23, 2021 The concept of machine learning generate best results in health care data, it also reduce the work load of health care industry. This algorithm potentially overcome the issues and find out the novel knowledge for development of medical date in health care industry. In this paper propose a new algorithm for finding the outliers using different datasets. Considering that medical data are analytic of mutually health problems and an activity. The proposed algorithm is working based on supervised and unsupervised learning. This algorithm detects the outliers in medical data. The effectiveness of local and global data factor for outlier detection for medical data in real time. Whatever, the model used in this scenario from their training and testing of medical data. The cleaning process based on the complete attributes of dataset of similarity operations. Experiments are conducted in built in various medical datasets. The statistical outcome describe that the machine learning based outlier finding algorithm given that best accurateness. Keywords: Machine learning algorithm Medical data analysis Outlier’s detection Similarity operations This is an open access article under the CC BY-SA license. Corresponding Author: R. Vijaya Kumar Reddy Department of IT Lakireddy Bali Reddy College of Engineering (Autonomus) Mylavaram, India E-mail: Vijayakumarr285@gmail.com 1. INTRODUCTION Outlier recognition is significant themes in data mining, the aim of discovery pattern that happen rarely as opposite to other data mining methods [1]. An outlier is derived significantly from inconsistent of a dataset [2]. The significance of outlier finding is in the sight of the truth so as to outliers can offer raw patterns and precious information about a dataset. Present research cover the fields of outlier discovery along with credit card fraud finding, network intrusion exposure, crime discovery, medical analysis, and detecting unusual parts in image processing [3]. Unsupervised outlier detection, is normally classified into distance-based [4], [5], density-based [6], [7], and distribution-based [3] procedures. This approach finds each data points are produced by a definite arithmetical model, but outliers do not accept this type of model. This method was preliminarily investigated by Knox and Ng [5]. In local information of the dataset differ to the global parameters. Density-based method was at first discussed by Breunig et al. [6]. Based on their local point density, local outlier factor is assign to every data point. The data point with a far above the ground for the local (LOF) value is described as an outlier. The clustering-based methods are unsupervised, they d not require any labeled training data, and their appearance in outlier discovery is restricted. Many real-world applications may come across dissimilar cases for a small set of objects are labeled as outliers to a certain class, but the greater part of data are unlabeled. Significantly improve the efficiency of outlier detection, little bit of proper knowledge is required [8]-[10]. So semi supervised methods to outlier