ML AND DL CROSS-DOMAIN ANALYSIS ASSESSING THEIR INFLUENCE IN VARIOUS DOMAINS

Abstract

Machine Learning (ML) and Deep Learning (DL) technologies are indispensable in the fields of solving complex problems across different application areas. In this study, the authors compare cross domain use of ML and DL algorithms across three industry sectors—healthcare, natural language processing (NLP), and network security. The pre-processing of publicly available benchmark datasets was performed on the data using Data Cleaning, Normalization, Feature Selection and Text Preprocessing methods. A total of 4 ML algorithms (Random Forest, XGBoost, LightGBM and CatBoost) and 4 DL algorithms (CNN, LSTM, GRU and Transformer) were tested on classification accuracy. The outputs show that the deep learning models generally outperform the machine learning algorithms with the highest accuracy in all the domains being achieved by Transformer and XGBoost being the best-performing machine learning algorithm. This research points out advantages and disadvantages of both strategies, and guides selection of appropriate algorithms in a variety of real-world applications.