Survey on establishing the optimal number of factors in exploratory factor analysis applied to data mining

Abstract

In many types of researches and studies including those performed by the sciences of agriculture and plant sciences, large quantities of data are frequently obtained that must be analyzed using different data mining techniques. Sometimes data mining involves the application of different methods of statistical data analysis. Exploratory Factor Analysis (EFA) is frequently used as a technique for data reduction and structure detection in data mining. In our survey, we study the EFA applied to data mining, focusing on the problem of establishing of the optimal number of factors to be retained. The number of factors to retain is the most important decision to take after the factor extraction in EFA. Many researchers discussed the criteria for choosing the optimal number of factors. Mistakes in factor extraction may consist in extracting too few or too many factors. An inappropriate number of factors may lead to erroneous conclusions. A comprehensive review of the state-of-the-art related to this subject was made. The main focus was on the most frequently applied factor selection methods, namely Kaiser Criterion, Cattell's Scree test, and Monte Carlo Parallel Analysis. We have highligthed the importance of the analysis in some research, based on the research specificity, of the total cumulative variance explained by the selected optimal number of extracted factors. It is necessary that the extracted factors explain at least a minimum threshold of cumulative variance. ExtrOptFact algorithm presents the steps that must be performed in EFA for the selection of the optimal number of factors. For validation purposes, a case study was presented, performed on data obtained in an experimental study that we made on Brassica napus plant. Applying the ExtrOptFact algorithm for Principal Component Analysis can be decided on the selection of three components that were called Qualitative, Generative, and Vegetative, which explained 92% of the total cumulative variance.

Classification of Scale Items with Exploratory Graph Analysis and Machine Learning Methods

“…On the other hand, studies on factor analysis using machine learning generally focused on factor retention (e.g. Goretzko & Bühner, 2020;Iantovics et al, 2019). Therefore, the results of the present study provide researchers with a reference point in using and selecting the most suitable machine learning method for their data structure to decide on which items will be included in which factors.…”

“…When studies on exploratory factor analysis using machine learning methods are examined, it can be seen that such studies generally focus on factor retention (e.g. Goretzko & Bühner, 2020;Iantovics et al, 2019). As a result of these studies, it has been reported that machine learning methods can generally be used with traditional methods.…”

Retrospective Study Regarding Orthodontic Retention Complications in Clinical Practice

“…For the verification of the normal distribution of percentages regarding the choice of an appropriate study design [26], we used the Shapiro-Wilk test (SW test) [27]. The selection of the SW test instead of the One-sample Kolmogorov-Smirnov test or the Lilliefors test was because Shapiro-Wilk test has the highest statistical power and works well with few data [28].…”

Exploratory Factor Analysis for Identifying Comorbidities as Risk Factors Among Patients with Cied

“…Based on the in-depth bibliographic study of the scientific literature, and a previous research [16], we took into account the following three rules to determine the number of extracted factors:…”

“…Identification of an inappropriate number of factors may lead to inaccurate conclusions. A previous article studied the importance of the total cumulative variance that should be explained by the selected most appropriate number of extracted factors [16]. It was considered that, at least, a minimum threshold of cumulative variance should be explained by the extracted factors that depends on the specificity of the research.…”