This research used the PSO algorithm to develop three new models, PSOGARCH, PSOEGARCH, and PSOGJR-GARCH, for improving business performance management. The tracking error methods are compared among the models in order to obtain a forecasting model with better performance. The three traditional time series models, GARCH, EGARCH, and GJR-GARCH, are used to undertake foreign exchange forecasting, and the results of these are compared to those of PSOGARCH, PSOEGARCH, and PSOGJR-GARCH models. The PSOGJR-GARCH model had the smallest error and the best forecasting ability, followed by the PSOEGARCH and PSOGARCH models, with the traditional GARCH models having the worst performance.