FEATURE INVESTIGATION IN A LEARNING PROCESS
Long-Shuh Lin, Hsiao‐Fan Wang · Journal of the Chinese Institute of Industrial Engineers · 2001
Obtaining sufficient and significant features of a system is a key issue on learning, which is correspondent to structural learning and parametric learning. In this study, we developed two functions to measure the completeness and significance of features for a system from probabilistic viewpoint. Then, a learning system which incorporates a multi-objective learning model and a learning procedure was proposed to effectively increase the information of a concerned system. Theoretical support was developed with computation analysis. The system has been applied to determine the features of traffic distribution for the central area of Taipei, Taiwan with satisfactory results.