基于ANN模糊协同方法半导体产业成本预测研究

针对半导体成本会随学习过程下降具有极大不确定性的特点，借用模糊协同思想，提出了一种ANN模糊协同智能的方法，对半导体单位成本进行预测。由代理应用单层感知机(ANN)拟合单位成本曲线；利用模糊交联法对代理的预测结果进行聚合，以缩小预测范围提高预测准确性；采用BPN算法对聚合结果去模糊化得到单位成本精确值。实验结果表明，MAE、MAPE、RMSE等各项指标都有显著提高。

Fuzzy Collaborative Method Based on ANN Semiconductor Industry Cost Prediction Research

Aiming at the problem that the cost of semiconductor decreases with the learning process and has great uncertainty, a ANN fuzzy collaborative intelligence method is proposed to predict the unit cost of semiconductor. a single layer perceptron(ANN) is used to fit the unit cost curve by agent. Fuzzy interaction method is used to aggregate the agent''s prediction results to reduce the prediction range and improve the accuracy of prediction. BPN algorithm was used to defuzzize the aggregation results to get the accurate unit cost. The results showed that MAE, MAPE and RMSE are improved significantly.