Self-organized and Unsupervised Information Integration Method by Combining Capacity and PCA
Jian‐Zhang Wu, Yanqing Li, Fengfeng Chen, Gleb Beliakov · ECONOMIC COMPUTATION AND ECONOMIC CYBERNETICS STUDIES AND RESEARCH · 2021
The Choquet capacity and integral is a widely accepted model to aggregate the multiple dimensional information mainly for its preeminent ability to flexibly represent and efficiently deal with the interaction phenomenon among the multiple correlative criteria/items.The PCA (Principal Component Analysis) is a popular approach to obtain the successive importance of multiple criteria as well as the correlations among them from the inter-correlated multiple dimensional data.In this paper, we combine some unsupervised capacity identification methods and PCA to constitute a self-organized information aggregation scheme, in which the PCA is used to generate the partial or comprehensive importance and correlations of the reduced (if necessary) multiple criteria, then adopting these preference information as the input, the capacity identification methods transform them into the competent capacities according to some rules and principles, and the final comprehensive integrated information is carried out by the Choquet integral accordingly.The main characteristic of this scheme is it only takes the raw unscaled data on the multiple criteria as the input and many meaningful aggregation analysis and decision aid results can be automatically output without the analyst or decision maker's supervision.The feasibility and practicability of the proposed models are demonstrated by using the QS World University Rankings data.