A comparison of multiple linear regression and random forest for community concern of youth and young adults survey

Abstract

The youth and young adults are an essential part of a community’s development. Therefore, an assessment of their concerns and related factors could help reflect the overall situation in the community. In this study, the community problems of concern to youth and young adults in three districts of Pattani province are addressed. The data were collected using a questionnaire consisting of 31 items for the problems of concern, and targeting 460 youth and young adults in the focus area. This study aimed to compare the performances of two methods to explore the related factors in the survey data. Those two methods are multiple linear regression (MLR), representing a conventional statistical method, and random forest (RF), representing a machine learning approach. In the results, the random forest regression models seemed superior to the multiple linear regression models in predictive performance and errors. The findings indicate that using RF for data analysis of survey results can be an alternative to a conventional approach in social sciences research.

Keywords

• modeling
• applied social science
• multiple linear regression
• random forest
• survey