In the second is about the ann ' s structure , one genetic arithmetic ( ga ) is used to chose the most logical number of connotative layer , so it can void the blindness of chose the number of connotative layer . because based on more logical net structure , the precision of forecasting is improved 第二部分研究神經網絡的結構問題,利用遺傳算法對網絡的隱含層個數進行尋優,從而避免了隱含層選擇的盲目性,使得預測在更加合理的網絡結構上進行,提高了預測的精度。