On the use of Sparse Principal Component Analysis and Robust: Selection Features of Maize Yield in Rural Tanzania
No Thumbnail Available
Date
2017
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
MANECH Pblications
Abstract
This paper has been motivated as a result of an existence of high dimensionality problem in maize yield. This means that an application of the Sparse Principal Component Analysis (SPCA) pattern recognition technique is unknown in selecting few consistent features and easier interpretation as opposed to classical PCA. This paper fulfills the existing knowledge gap in the context of Tanzania. A structure questionnaire was used to collect primary data from Mbozi and Mvomero Districts among small farming household in rural areas. The study was designed on the basis of hierarchical random sampling. The breakdown of facts was made by R-Statistical computing (version 3.3.2) whereas the findings were depicted using graphs and tables. The statistical estimates like percentage, mean and variance were also used. In line with SPCA, PCA and Robust PCA were also fitted for comparison purpose. Results showed 19 variables were condensed to six components explaining 63.7 per cent variations under PCA. Contrary to these findings, there were great improvements of the loadings, consistent and easier to interpret in each PC of the modified model (SPCA). However, the paper discovered that the Robust PCA condensed the p-variable to two PCs such that PC1 explained (81.0 per cent) variances. The study recommends the Sparse and Robustness as the best filtering techniques with reliable results as contrasted to the ordinary PCA.
Description
Article published by MANECH Publications in the Journal of Engineering Mathematics and Statistics Volume 1 Issue 1
Keywords
Classical Principal Components Analysis, Sparse Principal Component Analysis, Dimensionality Reduction, Robustness, Smallholder Farmers and Maize Yield
Citation
APA