Feature Selection: A Practitioner View

Goswami, Saptarsi; Chakrabarti, Amlan

doi:10.5815/ijitcs.2014.11.10

Cited by 50 publications

(22 citation statements)

References 36 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In paper [10], authors have discussed a feature selection technique based on clustering the coefficient of variations. As observed in paper [11], SPSS, which is a leading commercial tool for data mining by IBM recommends screening those features which has a low coefficient of variation. Another commercial tool SQL Server Analysis Service from Microsoft outlines the importance of entropy in finding interestingness or importance of an attribute [12].…”

Section: Related Workmentioning

confidence: 99%

An Univariate Feature Elimination Strategy for Clustering Based on Metafeatures

Goswami¹,

Chakraborty²,

Saha³

2017

IJISA

Self Cite

View full text Add to dashboard Cite

Abstract-Feature selection plays a very important role in all pattern recognition tasks. It has several benefits in terms of reduced data collection effort, better interpretability of the models and reduced model building and execution time. A lot of problems in feature selection have been shown to be NP -Hard. There has been significant research in feature selection in last three decades. However, the problem of feature selection for clustering is still quite an open area. The main reason is unavailability of target variable as compared to supervised tasks. In this paper, five properties or metafeatures like entropy, skewness, kurtosis, coefficient of variation and average correlation of the features have been studied and analysed. An extensive study has been conducted over 21 publicly available datasets, to evaluate viability of feature elimination strategy based on the values of the metafeatures for feature selection in clustering. A strategy to select the most appropriate metafeatures for a particular dataset has also been outlined. The results indicate that the performance decrease is not statistically significant.

show abstract

Section: Related Workmentioning

confidence: 99%

An Univariate Feature Elimination Strategy for Clustering Based on Metafeatures

Goswami¹,

Chakraborty²,

Saha³

2017

IJISA

Self Cite

View full text Add to dashboard Cite

show abstract

“…Conceptualized, data mining and knowledge discovery on the unstructured data is needed to be carried out (Goswami and Chakrabarti, 2014). But the choice of machine learning algorithm to solve a problem always depends on the size, quality, and nature of the data (Djam et al, 2011).…”

Section: Choice Of Predicting Modelmentioning

confidence: 99%

Symptomatic and Climatic Based Malaria Threat Detection Using Multilevel Thresholding FeedForward Neural Network

A.¹,

R.²

2017

IJITCS

View full text Add to dashboard Cite

Abstract-Recent worldwide medical research is focusing on new intelligence approaches for diagnosis of various infections. The sporadic occurrence of malaria diseases in human has pushed the need to develop computational approaches for its diagnoses. Most existing conventional malaria models for classification problems examine the dynamics of asymptomatic and morphological characteristics of malaria parasite in the thick blood smear, but this study examine the symptomatic characteristics of malaria parasite combined with effects of climatic factor which is a great determinant of malaria severity. The need to predict the occurrence of malaria disease and its outbreak will be helpful to take appropriate actions by individuals, World Health Organizations and Government Agencies and its devastating impact will be reduced. This paper proposed Feed-Forward Back-Propagation (FF_BP) Neural Network model to determine the rate of malaria transmission. Monthly averages of climatic factors; rainfall, temperature and relative humidity with monthly malaria incidences were used as input variables. An optimum threshold value of 0.7100 with classification accuracy 87.56%, sensitivity 96.67% and specificity 76.67% and mean square error of 0.100 were achieved. While, the model malaria threat detection rate was 87.56%, positive predictive value was 89.23%, negative predictive value was 92.00% and the standard deviation is 2.533. The statistical analysis of Feed-Forward BackPropagation Neural Network model was conducted and its results were compared with other existing models to check its robustness and viability.

show abstract

“…The objective of feature selection was to choose a subset of input variables by eliminating features, which are irrelevant or of no predictive information [20].…”

Section: Data Pre-processingmentioning

confidence: 99%

Data Mining Model for Predicting Student Enrolment in STEM Courses in Higher Education Institutions

Wanjau¹,

Okeyo²,

Rimiru³

2016

IJCATR

View full text Add to dashboard Cite

Educational data mining is the process of applying data mining tools and techniques to analyze data at educational institutions. In this paper, educational data mining was used to predict enrollment of students in Science, Technology, Engineering and Mathematics (STEM) courses in higher educational institutions. The study examined the extent to which individual, sociodemographic and school-level contextual factors help in pre-identifying successful and unsuccessful students in enrollment in STEM disciplines in Higher Education Institutions in Kenya. The Cross Industry Standard Process for Data Mining framework was applied to a dataset drawn from the first, second and third year undergraduate female students enrolled in STEM disciplines in one University in Kenya to model student enrollment. Feature selection was used to rank the predictor variables by their importance for further analysis. Various predictive algorithms were evaluated in predicting enrollment of students in STEM courses. Empirical results showed the following: (i) the most important factors separating successful from unsuccessful students are: High School final grade, teacher inspiration, career flexibility, pre-university awareness and mathematics grade. (ii) among classification algorithms for prediction, decision tree (CART) was the most successful classifier with an overall percentage of correct classification of 85.2%. This paper showcases the importance of Prediction and Classification based data mining algorithms in the field of education and also presents some promising future lines.

show abstract

Feature Selection: A Practitioner View

Cited by 50 publications

References 36 publications

An Univariate Feature Elimination Strategy for Clustering Based on Metafeatures

An Univariate Feature Elimination Strategy for Clustering Based on Metafeatures

Symptomatic and Climatic Based Malaria Threat Detection Using Multilevel Thresholding FeedForward Neural Network

Data Mining Model for Predicting Student Enrolment in STEM Courses in Higher Education Institutions

Contact Info

Product

Resources

About