A Principal Component Analysis (PCA)-based framework for automated variable selection in geodemographic classification

Liu, Yunzhe; Singleton, Alex; Arribas‐Bel, Daniel

doi:10.1080/10095020.2019.1621549

Cited by 34 publications

(19 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…6.1 Selecting the Appropriate Indicators Liu et al (2019) explain that the selection of variables must consider theoretical aspects, previous experiences, practicality, availability, and statistical and data properties. The latter includes eliminating outliers and poorly correlated indicators, which can compromise the cluster's structure and consistency (Kim, 2015).…”

Section: How To Reduce the K-means Shortcomings?mentioning

confidence: 99%

Time–Space Analysis of Multidimensional Phenomena: A Composite Indicator of Social Exclusion Through k-Means

et al. 2021

View full text Add to dashboard Cite

Composite Indicators are one-dimensional measurements that simplify the interpretation of multidimensional phenomena that facilitate public policies' elaboration. The literature on composite indicators is abundant, diversified, and inserted in practically all knowledge areas. Part of this literature aims to reduce uncertainties that propagate through the structure of the composite indicator during the process of normalization, weighting, and aggregation of indicators. Even if no composite indicator is exempt from criticism, the current literature is already sufficiently large and deep to guide researchers in constructing reliable composite indicators. However, most related works are concerned with representing multidimensional phenomena in time or space. Although some studies are interested in representing multidimensional phenomena that co-occur in time-space, the portion of the literature that addresses composite indicators is still not comprehensive, therefore leaving several open questions: What are the additional challenges in representing multidimensional phenomena in time-space? What methods can be used? Which method is most appropriate for this type of representation? What are the shortcomings of this method? How to reduce these shortcomings? This research aims at answering these questions in order to advance the time-space analysis of multidimensional phenomena. As a general contribution, the work presents a scheme of procedures that reduce subjectivities and uncertainties in the representations of multidimensional phenomena in time-space. As a specific contribution, it provides accurate and reliable information on the trajectory of social exclusion in the analyzed region. KeywordsComposite indicators • Multidimensional phenomena • Cluster analysis • Principal component analysis • Time-in-space analysis • Urban geography

show abstract

Section: How To Reduce the K-means Shortcomings?mentioning

confidence: 99%

Time–Space Analysis of Multidimensional Phenomena: A Composite Indicator of Social Exclusion Through k-Means

et al. 2021

View full text Add to dashboard Cite

show abstract

“…In this study, the above procedure was conducted for SLPs for 1-to 3-day lags and the PCs of each lag were used as model input. e PCA has been found as an efficient method for the selection of inputs from a large dataset in previous studies [57,58].…”

Section: Generation Of Model Inputsmentioning

confidence: 99%

Bayesian Regularized Neural Network Model Development for Predicting Daily Rainfall from Sea Level Pressure Data: Investigation on Solving Complex Hydrology Problem

Jabbar

Zahra

et al. 2021

Complexity

View full text Add to dashboard Cite

Prediction of daily rainfall is important for flood forecasting, reservoir operation, and many other hydrological applications. The artificial intelligence (AI) algorithm is generally used for stochastic forecasting rainfall which is not capable to simulate unseen extreme rainfall events which become common due to climate change. A new model is developed in this study for prediction of daily rainfall for different lead times based on sea level pressure (SLP) which is physically related to rainfall on land and thus able to predict unseen rainfall events. Daily rainfall of east coast of Peninsular Malaysia (PM) was predicted using SLP data over the climate domain. Five advanced AI algorithms such as extreme learning machine (ELM), Bayesian regularized neural networks (BRNNs), Bayesian additive regression trees (BART), extreme gradient boosting (xgBoost), and hybrid neural fuzzy inference system (HNFIS) were used considering the complex relationship of rainfall with sea level pressure. Principle components of SLP domain correlated with daily rainfall were used as predictors. The results revealed that the efficacy of AI models is predicting daily rainfall one day before. The relative performance of the models revealed the higher performance of BRNN with normalized root mean square error (NRMSE) of 0.678 compared with HNFIS (NRMSE = 0.708), BART (NRMSE = 0.784), xgBoost (NRMSE = 0.803), and ELM (NRMSE = 0.915). Visual inspection of predicted rainfall during model validation using density-scatter plot and other novel ways of visual comparison revealed the ability of BRNN to predict daily rainfall one day before reliably.

show abstract

“…For example, the clustering algorithms that underpin the development of the classifications will identify the 'best' groupings of the small-area geographies based on a mathematical optimisation process; however, the areas could have been grouped differently based on different input variables [37,43]. Therefore, it is critical that input variables are selected for their ability to generate relevant and meaningful groups, which has led to renewed exploration of variable selection procedures [44]. Yet, despite this, no examples in either the academic or the commercial literature have been found which suggest that any methods are employed at this stage for evaluating the contextual relevance or appropriateness of the input variables chosen.…”

Section: Proposed Alternatives To the Traditional Geodemographic Classification Development Frameworkmentioning

confidence: 99%

Local and Application-Specific Geodemographics for Data-Led Urban Decision Making

et al. 2021

View full text Add to dashboard Cite

This work seeks to introduce improvements to the traditional variable selection procedures employed in the development of geodemographic classifications. It presents a proposal for shifting from a traditional approach for generating general-purpose one-size-fits-all geodemographic classifications to application-specific classifications. This proposal addresses the recent scepticism towards the utility of general-purpose applications by employing supervised machine learning techniques in order to identify contextually relevant input variables from which to develop geodemographic classifications with increased discriminatory power. A framework introducing such techniques in the variable selection phase of geodemographic classification development is presented via a practical use-case that is focused on generating a geodemographic classification with an increased capacity for discriminating the propensity for Library use in the UK city of Leeds. Two local classifications are generated for the city, one a general-purpose classification, and the other, an application-specific classification incorporating supervised Feature Selection methods in the selection of input variables. The discriminatory power of each classification is evaluated and compared, with the result successfully demonstrating the capacity for the application-specific approach to generate a more contextually relevant result, and thus underpins increasingly targeted public policy decision making, particularly in the context of urban planning.

show abstract

A Principal Component Analysis (PCA)-based framework for automated variable selection in geodemographic classification

Cited by 34 publications

References 19 publications

Time–Space Analysis of Multidimensional Phenomena: A Composite Indicator of Social Exclusion Through k-Means

Time–Space Analysis of Multidimensional Phenomena: A Composite Indicator of Social Exclusion Through k-Means

Bayesian Regularized Neural Network Model Development for Predicting Daily Rainfall from Sea Level Pressure Data: Investigation on Solving Complex Hydrology Problem

Local and Application-Specific Geodemographics for Data-Led Urban Decision Making

Contact Info

Product

Resources

About