Krzysztof Siminski
Abstract:
Incomplete data are common and require special techniques. The essential techniques are: marginalisation, imputation, and rough sets. The paper presents the imputation by inversion of the neuro-fuzzy system. First the neuro-fuzzy systems is trained with complete data. Next the system is inverted and the missing values are imputed. The complete and imputed data are used to train the final neuro-fuzzy system. The technique is limited to data items with one missing value. The paper is accompanied by numerical examples and statistical verification.
Reference:
Krzysztof Siminski, Memetic Neuro-Fuzzy System with Big-Bang-Big-Crunch Optimisation, [chapter in] Man--Machine Interactions 4, Springer International Publishing, 2016, pp. 583-592.
Bibtex Entry:
@incollection{id:Siminski2016Memetic,
title={Memetic Neuro-Fuzzy System with {B}ig-{B}ang-{B}ig-{C}runch Optimisation},
author={Siminski, Krzysztof},
booktitle={Man--Machine Interactions 4},
pages={583--592},
year={2016},
publisher={Springer International Publishing},
abstract = {Incomplete data are common and require special techniques.
The essential techniques are: marginalisation, imputation, and rough
sets. The paper presents the imputation by inversion of the neuro-fuzzy
system. First the neuro-fuzzy systems is trained with complete data.
Next the system is inverted and the missing values are imputed. The
complete and imputed data are used to train the final neuro-fuzzy system.
The technique is limited to data items with one missing value. The
paper is accompanied by numerical examples and statistical verification.},
}