Disjoint principal component analysis by constrained binary particle swarm optimization
This work addresses a specific computational challenge in data analysis for researchers, but it is incremental as it builds on existing disjoint PCA methods.
The paper tackles the problem of disjoint principal component analysis by proposing a method that uses constrained binary particle swarm optimization to find disjoint components as linear combinations of subsets of variables, with numerical results confirming solution quality.
In this paper, we propose an alternative method to the disjoint principal component analysis. The method consists of a principal component analysis with constraints, which allows us to determine disjoint components that are linear combinations of disjoint subsets of the original variables. The proposed method is named constrained binary optimization by particle swarm disjoint principal component analysis, since it is based on the particle swarm optimization. The method uses stochastic optimization to find solutions in cases of high computational complexity. The algorithm associated with the method starts generating randomly a particle population which iteratively evolves until attaining a global optimum which is function of the disjoint components. Numerical results are provided to confirm the quality of the solutions attained by the proposed method. Illustrative examples with real data are conducted to show the potential applications of the method.