A Comprehensive Study of SOMs, iSOMs, and Hybrid SOMs for Complex Data

With the help of data-driven insights and decision-making, machine learning has emerged as a transformational force across various areas. Understanding complex datasets and facilitating interpretability are key visualization functions. Here, we present the potential of Self Organizing Maps (SOMs), Incremental Self Organizing Maps (iSOMs), and Hybrid SOMs as revolutionary tools for data visualization. These methods can potentially solve the problems brought on by high dimensional and diverse data. SOMs, a subclass of unsupervised neural networks, are well known for their capacity to preserve topological structure while mapping high dimensional input onto smaller level grids. Our study reviews the previous techniques, while exploring the new uses of SOMs for data exploration, dimensionality reduction, and clustering. We highlight their versatility in revealing buried patterns and abnormalities as we explain their adaptation to diverse data sources, including textual, image, and genomic data. This study illustrates the potential of traditional SOMs, iSOMs, and hybrid SOMs as adaptable tools for data visualization and analysis by providing a comprehensive overview of each type of SOM. We demonstrate their efficacy in numerous application domains through different datasets and experimental findings, opening the path for improved interpretability and insights from complicated data. These methods offer researchers and practitioners innovative solutions for tackling the challenges posed by modern data analytics.