15160

13633

كارشناسي ارشد

بهينه سازي سيستم ها

اصفهان : دانشگاه صنعتي اصفهان

1398

يازده، 137 ص، مصور، جدول، نمودار

مهدي خاشعي

مهدي بيجاري، رضا حجازي

1398/07/19

كتابنامه

مهندسي صنايع

مهندسي صنايع و سيستم ها

1398/07/20

2568341

AbstractZinc is one of the most important and strategic basic metals which has a special place in the Iran industry aswell as the world Zinc price fluctuations and its frequent price changes in different periods are a reflection ofthe world economic conditions One of the main concerns facing zinc industry activities is to find the futureprices and to easier make trading decisions Since the future planning is required to be estimated theforthcoming Mining companies and factories that are active in the extraction and production of zinc alwaysface risks and fluctuations for planning economic analysis of projects and reviewing new investment fordevelopment etc due to market price changes These fluctuations and uncertainty about the future causesevere losses for companies and even leaving the competitive market and the reluctance of others to enter thefield Studies show that the price of zinc in Iran is strongly influenced by its worldwide price The purpose ofthis thesis is to provide a model to forecast the global price of zinc Due to the existence of multipleforecasting methods that each has advantages and disadvantages this thesis presents a hybrid forecasting thathas the capability of linear modeling of classical statistical methods as well as the power of nonlinearmodeling of artificial neural networks Initially the list of primary effective variables is identified based onthe literature of the subject then the list of final variables is determined by examining the amount ofexplanation and linear and nonlinear correlations between variables Following the general principles ofmultilayer perceptron neural network design and multivariate linear regression different hybrid models havebeen designed and implemented using these two methods and the results of these models are compared witheach other By applying the proposed hybrid model a 26 87 and 6 29 improvement based on the rootmean square error in forecasting of test data set is obtained compared to multivariate linear regressionmodel and multilayer perceptron neural network respectively The results of this thesis indicate that theproposed hybrid model has good accuracy and performance in forecasting the global price of zinc Keywords Forecasting Global price of zinc Classical statistical models Computational intelligent tools Hybrid forecasting models

مهدي خاشعي

مهدي بيجاري، رضا حجازي