Prediction of Unemployment Rates in the United States by K-Nearest Neighbor Regression Analysis

Abstract

Unemployment remains a pervasive challenge on the global stage, bearing significant social, psychological, and economic ramifications. This issue touches the core of human existence, influencing not just the fundamental necessities of life but also personal aspirations and leisure activities. For nations on the cusp of economic growth, like the United States, unemployment is a formidable obstacle to achieving their growth aspirations. Addressing this concern requires a forward-looking approach, where predictions about future unemployment rates are made based on accessible data. This study embarks on such a mission, focusing on the unemployment rates in the United States. It employs the k-nearest neighbor regression (kNNR) to make these estimations. To enhance the precision of these forecasts, we have assembled a new dataset, including factors presumed to impact the behavior of unemployment. Encompassing factors believed to have a bearing on unemployment dynamics. To assess the effectiveness of the kNNR, we have juxtaposed its performance with that of another prominent machine learning contender: linear regression and decision trees. The empirical outcomes are telling with a coefficient of determination (R2 value of 0.936; the kNNR not only showcased its predictive prowess but also outshone the other algorithms in the fray. These findings underscore the potential of the kNNR algorithm as a potent tool in unemployment rate prediction endeavors.