New Index of Significance for Inputs Selection
Joanna Chimiak–Opoka · 2003
The reduction of dimensionality should be performed before applying any modelling tool. In the process of inputs selection we need an index to order the inputs according to their significance. In the paper a new index of significance is presented and the algorithm of its estimation in full-dimensional input-space is proposed. The algorithm subdivides space into cells and computes membership function for each of them. The experimental results, which were obtained using the proposed method, are very promising. Finally a benchmark example demonstrates how the process of inputs selection is applied.