The financial market is a rapidly growing industry where derivatives are popular among investors. The rapid growth of options trading leads to the development of various option pricing theories and models. Through them, Longstaff and Schwartz improved the Monte Carlo model in 2001. The improved least square Monte Carlo simulation (LSM) is widely used in pricing American options. This paper aims to compare two estimation methods in pricing American options, namely the Least Square Monte Carlo Algorithm and the Binomial Tree Model, and detect which model better estimates the accuracy of the operation. In a refinement of using the Generalized Autoregressive Conditional Heteroskedasticity (GARCH) model to measure the volatility, an empirical comparison of the models using the Copper future contract is conducted. It is found that the binomial tree method can more accurately predict the American options prediction problem, and applying it to the pricing of copper can not only improve the market but also provide a more reasonable and rapid pricing basis for its subsequent development.