Advancements in Household Classification Using Multiclass Decision Forest: A Case Study of the Kut Bak District, Sakon Nakhon Province

Abstract

This study examines the classification of household poverty in Kut Bak District, Sakon Nakhon Province, Thailand, using advanced machine learning techniques, specifically the Multiclass Decision Forest algorithm. The research pursued two main objectives: (1) to develop a robust model for classifying households into distinct poverty levels, and (2) to identify key socio-economic factors that significantly influence poverty classification. The dataset comprised 302 households, representing various socio-economic strata within the district. The study population consisted of 686 households, with the sample selected through a participatory process involving local community leaders and government officials. Nineteen socio-economic features were included in the model, such as household income, expenditures, and participation in local development projects. Households were classified into four poverty levels: extremely poor, moderately poor, marginally poor, and above poor. Model performance was evaluated using confusion matrices, yielding an overall accuracy of 94.4%, including perfect (100%) accuracy in classifying the extreme categories (extremely poor and above poor). The analysis further revealed that household income and participation in local projects were the most influential determinants of poverty levels. Specifically, households with low income and minimal project participation were most likely to fall into the “extremely poor” category, whereas those with higher income and multiple income sources were typically classified as “above poor.” The findings provide important insights into the socio-economic dynamics of poverty in rural Thailand and offer practical implications for designing targeted poverty alleviation strategies. More broadly, the study demonstrates the potential of machine learning approaches to address complex socio-economic challenges, thereby contributing to the field of development studies.