Exploratory Data Analysis for AI-Based Channel Selection for Stable Coastal Maritime Wireless Communications in Emergency Situations

Exploratory data analysis (EDA) plays an important role in various areas of research. In scenarios where reliable communication is essential, such as maritime environments, EDA can play a significant role in optimizing decision-making processes. In this paper, we perform EDA on a maritime communication dataset to enhance channel selection between Wi-Fi and LTE networks, a critical task for ensuring efficient communication based on radio availability. Using various feature selection techniques such as correlation matrix, SelectKBest, Recursive Feature Elimination (RFE), Chi-Square test, and Variance Threshold, key features are identified. The results reveal that features like TX_CCQ (Client Connection Quality for transmitted data) and RX_CCQ (Client Connection Quality for received data) are consistently significant in predicting network selection. EDA techniques demonstrate how there are marginal improvements in the accuracy for classification performance of Random Forest and LSTM (Long short-term memory) models.