基于混沌SVM与ARIMA的工程造价组合预测

为了解决工程造价预测的时效性问题，针对传统线性时间序列预测模型可靠性不高的缺点，引入混沌相空间重构和支持向量机技术，并将两者耦合组成一种非线性预测模型，再利用ARIMA在整体线性趋势预测方面的优越性，对非线性模型进行修正。混沌SVM和ARIMA预测构成组合模型的两个子过程，将两个子过程的预测结果综合平均即可得到最终预测结果。经实例计算，组合模型比最大Lyapunov指数、ARIMA和只将相空间重构与SVM进行耦合的方法拟合效果好，预测精度高，证明其的确具有线性趋势拟合和非线性波动拟合的双优势。

Combination Model for Predicting Construction Cost Based on Hybrid Chaos SVM and ARIMA

To solve the timeliness problem of construction cost prediction, aiming at low reliability defect of traditional linear time series prediction model, a hybrid model is proposed by coupling Chaos phase space reconstitution with SVM into a nonlinear prediction technique and utilizing ARIMA for its advantage in predicting overall linear trend as a modification for nonlinear prediction. Chaos-SVM and ARIMA are two sub procedures of this hybrid model, and final result is average of separate prediction outcomes. Case study shows that this model has better curve fitting and a higher accuracy than forecasting model based on largest Lyapunov index, ARIMA and Chaos-SVM only, which proves it indeed has superiority in both linear trend fitting and nonlinear curve fitting.