The ability to predict the impact of concurrent running queries on the behavior of individual queries is crucial to resource management, especially in the emerging database-as-a-service markets. Machine learning techniques have recently been applied for predicting query performance. This paper proposes the use of population-based metaheuristics (PBMHs) method for simultaneous feature selection and parameter optimization of Support Vector Regression (SVR), for increasing the prediction accuracy and capability of generalization. The proposed method is implemented with three different variations of PBMH, and their performance are compared. The authors evaluate their method using the TPC-DS benchmark, and compare the results obtained under dynamic workloads and static workloads in order to investigate the capability of generalization. One of the proposed methods produces the best and on average, 81.13% and 95.20% prediction accuracy, respectively, under the dynamic and static workloads. Therefore, the proposed method is valuable for concurrent query performance prediction.