[Paper Review] Feature Selection for Gender Classification in TUIK Life Satisfaction Survey

Adil Çoban, İlhan Tarımer|arXiv (Cornell University)|Jul 12, 2018

Psychological Well-being and Life Satisfaction7 references3 citations

TL;DR

This study proposes a feature selection approach to improve gender classification in the TÜİK Life Satisfaction Survey using attribute selection algorithms and machine learning models. By applying Chi2, Information Gain, and ReliefF to rank features, the best performance—73% accuracy—was achieved with Random Forest using only three selected attributes.

ABSTRACT

As known, attribute selection is a method that is used before the classification of data mining. In this study, a new data set has been created by using attributes expressing overall satisfaction in Turkey Statistical Institute (TSI) Life Satisfaction Survey dataset. Attributes are sorted by Ranking search method using attribute selection algorithms in a data mining application. These selected attributes were subjected to a classification test with Naive Bayes and Random Forest from machine learning algorithms. The feature selection algorithms are compared according to the number of attributes selected and the classification accuracy rates achievable with them. In this study, which is aimed at reducing the dataset volume, the best classification result comes up with 3 attributes selected by the Chi2 algorithm. The best classification rate was 73% with the Random Forest classification algorithm.

Motivation & Objective

• To reduce dataset dimensionality in the TÜİK Life Satisfaction Survey while preserving classification performance.
• To evaluate the effectiveness of multiple feature selection algorithms (Chi2, Information Gain, ReliefF) in identifying relevant attributes for gender classification.
• To compare classification accuracy across different algorithms (Naive Bayes, Random Forest) using selected features.
• To identify the optimal combination of feature selection method and classifier for gender prediction in survey data.

Proposed method

• Constructed a new dataset from the TÜİK Life Satisfaction Survey focusing on overall satisfaction attributes.
• Applied three attribute selection algorithms—Chi2, Information Gain, and ReliefF—to rank features based on their relevance to gender.
• Used the top-ranked features from each algorithm as input for classification with Naive Bayes and Random Forest.
• Evaluated performance using classification accuracy as the primary metric.
• Selected the best-performing feature subset based on accuracy and number of features used.
• Optimized the process by comparing the trade-off between feature count and classification accuracy.

Experimental results

Research questions

• RQ1 Which feature selection algorithm (Chi2, Information Gain, ReliefF) yields the highest gender classification accuracy?
• RQ2 What is the optimal number of features required to achieve high classification performance in this dataset?
• RQ3 How do the classification models (Naive Bayes and Random Forest) perform when trained on selected features compared to full feature sets?
• RQ4 Can a minimal set of features maintain high gender classification accuracy while reducing dataset complexity?

Key findings

• The Chi2 feature selection algorithm produced the best classification performance, achieving 73% accuracy with only three selected features.
• Random Forest outperformed Naive Bayes, delivering the highest accuracy among the tested classifiers.
• The combination of Chi2 selection and Random Forest classification achieved the optimal balance between feature count and accuracy.
• The study demonstrated that gender classification is feasible using a small, carefully selected subset of survey attributes.
• The number of selected features significantly impacted performance, with diminishing returns beyond a certain threshold.
• The results suggest that attribute selection can effectively reduce dataset size without sacrificing classification quality.