認購權證是選擇權的一種，提供投資人套利、避險等多樣化選擇。傳統選擇權訂價模式為一複雜之理論，訂價模式有許多限制，與實務上差距有待克服，因此本研究嘗試使用類神經網路建立認購權證評價模式。為避免差異，以Black-Scholes模式中，五項影響權證價格之因子為輸入變數，分別以倒傳遞網路與半徑式函數網路建立模式，並藉差異分析找出可改善學習績效之變數。由於各項風險係數真有模糊特性，在認購權證的操作策略與避險部位上，採用類神經模糊技術來建構，由結論中可得知建制認購權證智慧型系統是可行的。

Warrant is a type of call option. It provides people with multiple choice in speculate behavior contain arbitrage and hedging. Traditional option pricing model was a complex theory, and had a lot of limitation and assumption wait for overcome. This study tries to use artificial neural network to build option-pricing model for warrant. In Black-Scholes pricing model, there was five variables impact the option price that we take to be the input variable in artificial neural network, both Back - Propagation Network (BPN) and Radial Basis Function Network (RBFN) are used. Base on the difference analysis, we find out another variable that can improve learning efficiency and affectivity. The reason why using NeuroFuzzy on warrants operation strategy and hedging position is that the hedging coeffi­cient had fuzzy characteristic. However, NeuroFuzzy technology can take a turn for Artificial Neural Network can’t do.