计算机超频结果测量值的预测算法

Prediction Algorithm for Result's Measurements of Computer's Overclocking

摘要:

针对计算机超频结果测量值的估计问题,提出一种优化的LS-SVM回归模型解决方案；首先对CPU和显示卡超频的参数设置及主要结果进行了分析,并探讨了常用预测算法的不足,据此选择LS-SVM回归模型进行超频结果测量值的预测；然后设计了一种LCQPSO算法寻找模型参数的最优值,使模型的预测精度和泛化能力得以提高；经采集AMD FX-8350 CPU的50组超频样本验证,该算法的预测误差比RBF神经网络降低了80%,表明了该算法的有效性。

Abstract:

Aim at the estimation problem of the result values of computer’s overclocking, a solution to optimize LS-SVM regression model is proposed. First the parameter settings and its main results of CPU and graphics card’s overclocking are analyzed, and the deficiency of the commonly used prediction algorithm is discussed, for this the LS-SVM regression model is selected to predict the result’s measurements of overclocking. Then a LCQPSO algorithm is designed to find the optimal values of parameters and the model prediction accuracy and generalization ability is improved. Verified by collect 50 overclocking samples of AMD FX-8350, the prediction error of the algorithm is 80% lower than that of RBF neural network. the results show the effectiveness of the algorithm.