Determination of households benefits from subsidies by using data mining approaches

ABSTRACT Poverty, known as a widespread economic and political challenge (specifically at the times of crisis, like COVID-19), is a very complicated problem, which many countries have been trying for a long time to eradicate. Cash-subsidy allocation procedure using traditional statistical vision is the famous approach, which articles have targeted. Inefficiency of these solutions besides the fact that a pair of households with exact same situation will not be existing leads us to inadequacy and inaccuracy of these methods. This study, by putting data mining and machine learning (as well-known majors in IT and computer Science) visions together, draws a path to overcome this challenge. For this aim, the social, income and expenditure dimensions of a dataset are surveyed from 18885 households considered to measure the population poverty ratio (a fuzzy look at on their eligibility). In respect to the different experimental mode, the effective features are being filtered to use in FCM algorithm in order to determine to what extend the households in the poor or wealthy. Moreover, Genetic Algorithm displays its efficiency in the role of optimizer. Finally, the evaluation results show more accurate outcomes from the feature selection technique (on normalized data) and get the optimized clusters.