Prediction of Properties of Unknotted Spliced Ends of Yarns Using Multiple Regression and Artificial Neural Networks. Part II: Verification of Regression Models
Research and development
Authors:
- Lewandowski Stanisław
Institute of Textile Engineering and Polymer Materials, University of Bielsko-Biala, Bielsko-Biała, Poland - Drobina Robert (j/w)
 Full text | Abstract: A verification of regressive models based on artificial neural networks and multiple regression analysis was carried out. The analysis of the results obtained showed that artificial neural networks realizing regressive operations are useful for identifying the character of changes of additive quantities, in particular geometric dimensions and strength parameters. However, they are not suitable for identifying non-additive features, represented by tangling as well as teaseling. In this case, better predictive possibilities are provided by models based on multiple regression. |
Tags: layer diagrams, surface diagrams, learning quality, validation quality, testing quality, learning error, validation error, testing error, correlation.