Abstract: Because the steel price is nonlinear and its factor is difficult to be determined,in the forecast analysis,the tranditional method can only analyze the steel price with small amount of data,which leads to low accuracy of predection and the slow speed.In the big data era,memory computing in recent years has been a reseach hotpoint,and the requrement for timely data processing gets larger and larger.Based on the memory computing,the steel prices,production,inventory,and GDP data from 2002 to 2010,were used to build the prediction model,Bayessan forcasting model,ARMA model,support vector machine model and BP neural network model to forecast the steel prices.The simulation results show that the prediction model based on the memory not only has fast speed and high accuracy,but also shows the prices real timely.It provides a strong basis for enterprises to make decision on market reaction.

Key words: Big data,In-memory computing,Bayes,ARMA,Neural networks