Comparison of ant colony optimization and genetic algorithm methods for scale reduction

Measurement tools, which play a major role in making important decisions about individuals, especially in the fields of education and psychology, often consist of many items. In studies where multiple variables are measured together, the number of items increases even more, leading to respondent fatigue or incomplete responses. These problems have led researchers to shorten measurement instruments. The main purpose of this study is to compare Ant Colony Optimization and Genetic Algorithm methods used for scale shortening in item selection. Within the scope of the study, the short forms of two different scales obtained through Ant Colony Optimization and Genetic Algorithm were examined in terms of reliability values, explained variance ratios, factor loadings, model-data fit, and the relationship between total scores in the short form and total scores in the long form. When the results obtained in the study are evaluated, both methods produced short forms with high reliability values and good fit indices. Additionally, there is a highly statistically significant relationship between these short forms and the long forms. When the explained variance ratio is examined, the variance ratio explained by the short forms created with the Genetic Algorithm is almost the same as that of the long form, while the variance ratio explained by the short forms created with Ant Colony Optimization is higher than the variance explained in the long form.