Economic Indicators Selection for Property Crime Rates using Grey Relational Analysis and Support Vector Regression

Features selection is very important in the multivariate models because the accuracy of forecasting results produced by the model are highly dependent on these selected features. The purpose of this study is to propose grey relational analysis and support vector regression for features selection. The features are economic indicators that are used to forecast property crime rate. Grey relational analysis selects the best data series to represent each economic indicator and rank the economic indicators according to its importance to the property crime rate. Next, the support vector regression is used to select the significant economic indicators where particle swarm optimization estimates the parameters of support vector regression. In this study, we use unemployment rate, consumer price index, gross domestic product and consumer sentiment index as the economic indicators, as well as property crime rate for the United States. From our experiments, we found that the gross domestic product, unemployment rate and consumer price index are the most influential economic indicators. The proposed method is also found to produce better forecasting accuracy as compared to multiple linear regressions.