| dc.description.abstract | The rapid development of computer technology has led to the accumulation of large
amounts of high-dimensional data, creating challenges in analysis and visualization.
Data visualization plays an important role in revealing patterns and relationships
between variables, but the complexity of high-dimensional data often causes several
problems, such as overlap, uneven data density, and difficult separation between
clusters. Multidimensional Scaling (MDS) technique is one of the commonly used
methods for dimensionality reduction, but it still has limitations in maintaining
global structure and handling non-ideal data distribution. Therefore, this study
proposes an MDS-based optimization formulation that aims to improve the quality
of data visualization by minimizing distance distortion, improving local
distribution, and sharpening the separation between clusters. This approach is
developed based on a review of MDS+, MSSPD, Geometric MDS, and UAMDS
techniques, which have demonstrated advantages in handling high-dimensional
data. With this optimization formulation, data representation in a low-dimensional
space is expected to be clearer, more informative, and adaptive to data noise and
the addition of new data. In addition, this method can be applied in various fields
that require complex data analysis, such as big data visualization, image processing,
and data-based decision making. The results of this research are expected to make
a significant contribution to the development of more efficient and accurate data
visualization techniques, so as to improve data exploration and interpretation more
optimally. | en_US |
