Multi-objective optimization utilizing cluster analysis applied to dimensional transposed problems

Más información sobre el libro

This work addresses the significance of multi-objective optimization in modern information processing and analysis, highlighting the limitations of existing methods in handling large, complex tasks efficiently. It introduces a novel optimization concept that utilizes data domain transformations and cluster analyses to tackle multi-objective optimization problems. The approach simplifies the transposition of extensive, high-dimensional data models into low-dimensional, uniform equivalents within an independent framework. This framework is optimized for data similarity conservation, ensuring that the semantic relationships among data items are maintained, while also achieving low runtime complexity—where model size increases linearly correspond to runtime growth, despite considering all data relations. The cluster analysis is performed using an enhanced k-Means algorithm, designed to efficiently group large datasets into numerous clusters with linear time complexity. The application of these components to generic segmentation and pattern recognition tasks demonstrates the concept's effectiveness, yielding high-quality results and low runtimes. The abstracted components and their extensions are rigorously tested through full factorial design experiments with artificial, scalable data models, allowing for the determination of valid parameter ranges, assessment of result quality and runtime, and ensuring repeatable and comparable