A Comparison of Self-organising Maps and Principal Components Analysis

Abstract

This paper attempts to compare self-organising maps (SOM) and principal components analysis (CPA) by applying them to the marketing construct ‘retail store personality’. Data were collected for the retail store personality construct via a validated scale from previous studies that had used the mall intercept technique. A total of 367 people responded, of whom 353 were found to be valid for data analysis. Data were analysed using CPA and SOM; both methods gave comparable clustering results, although the results for SOM were quite conclusive. In addition, we found that SOM complemented PCA by providing visual clustering results far superior to those of PCA. SOM can be used to further analyse PCA data using visual clustering features; both could be used in tandem. Although SOM have been used in a number of studies in marketing, this is the first paper to compare PCA and SOM on terms of application to the marketing construct ‘retail store personality’.

Combining self-organizing maps and biplot analysis to preselect maize phenotypic components based on UAV high-throughput phenotyping platform

“…Finally, we ran SOM with the following parameters according to the guidelines for building a SOM reported by Das [51] and Wendel [52] : SOM size: 15 × 7; 5000 iterations; learning rate: 0.05, hexagonal topology and Gaussian neighborhood function. We resolved the clusters derived from the SOM map into a set of clustering rules by using the rpart [53] (version 4.1-15) and rpart.plot [54] (version 3.0.7) R package and evaluated clustering quality.…”

KPCA over PCA to assess urban resilience to floods

“…To measure the index FRI, the authors propose employing the extracted principal components to construct composite indices. [11] suggest using an intermediate resilience indicator IRI that correspond to each principal component to FRI. In contrast to the standard PCA, the extraction of a number of kernel principal components can exceed the input dimensionality p and depend to the chosen kernel function.…”

“…In practice, the proposed method does not yield the principal component axes, but the obtained eigenvectors can be understood as projections of the data onto the principal components. Unlike linear PCA method as proposed by [11], those eigenvectors already are the data points projected and can be employed directly to calculate FRI. In our case, the intermediate resilience indicators are given by eigenvectors of the centered kernel matrix that correspond to the largest eigenvalues.…”

“…The main purpose of PCA is the analysis of data to identify patterns that reflect the data. PCA is extensively applied to reduce the number of indicators highly correlated in a multivariate dataset to a smaller set of intermediate indicators [11]. In other words, PCA aims to find the axes with maximum variances along which the data is most spread.…”

Considering context and dynamics: A classification of transit-orientated development for New York City

“…The resulting map allows a graphical presentation of the data that can be easily interpreted by map-readers, which can be further classified by the machine learning techniques designed for low dimensionality (Bara et al, 2018;Spielman and Folch, 2015;Natita et al, 2016). Numerous studies have highlighted the utility of SOM for visualising complex, nonlinear statistical relationships within high-dimensional data (Yin, 2008;Bação and Lobo, 2010;Das et al, 2016;Miljković, 2017). The method is suitable for this application given the multiplex of measures assembled.…”