GENETIC TESTING FOR EARLY DETECTION AND PREVENTION OF HEREDITARY DISORDERS

Abstract

This research aims at assessing the efficacy of genetic testing in the early diagnosis and prevention of hereditary ailments Such prospects can be realize with the aid of modern machine learning algorithms. Using a set of genetic disorder tests as the data, a number of models, such as Auto_ViML – an automated machine learning model, and RandomForestClassifier, are deployed and tested to classify possible presence of genetic disorders. In order to overcome the issues these different classes pose as a large imbalance in the number of instances between the classes, we use SMOTE or the Synthetic Minority Over-sampling Technique in order to counterbalance the classes and hence make the calculations and the overall resultant models more accurate. This step is important in managing the given skewed data set characteristic to genetic disorders that more often possess fewer positive samples than negative ones. Also, for the purpose of explaining the models we employ LIME method that allows for the local model-agnostic explanation and provides an insight into how these black-box methods make decisions. The use of LIME allows the results of the machine learning models to be interpretable by the physicians, hence making them to trust the results of the models and or implement them into their practice. This paper emphasizes the importance of this feature to make the system more acceptable among practitioners who have to explain diagnoses and treatment plans to the patients. The findings revealed the prospects of automation in improving the conduct of screening for genetic disorders. Combining more sophisticated machine learning instruments with interpretability methodologies, our solution enables efficient detection of patients’ condition changes and contributes to their better health outcomes due to timely interventions and more precise treatment plans. The results call for the further integration of genomic tests and complex machine learning approaches to derive precise models that are implementable in clinical settings while being easy to explain. KEYWORDS: Network Intrusion Detection, UNSW-NB15, CIC-IDS2017, Packet Capture (PCAP), Machine Learning, Data Preprocessing, Feature Selection