In this study, molecular weight control studies and the development of advance control were carried out. In-line measurement device such as Gel Permeation Chromatography (GPC) has become available, however, it is very expensive and the results always posses substantial time-delayed measurements from analytical laboratory measuring devices. A method of predicting molecular weight performance during polymerization process was proposed using neural network system. A neural network model was developed to predict leading moments of molecular weight using backpropagation algorithm of neural networks system for Methyl methacrylate (MMA) polymerization. Plant input and output were simulated from the first principle model for MMA polymerization and then been utilized to train multilayer neural network system. Process inputs such as reaction temperature, monomer conversion and initiator concentration were the main variables affecting properties of molecular weight averages. A neural network model was generated from the training process after it successfully learned the relationship between process inputs and product outputs. This neural network model was applied when predicting molecular weight of Polymethyl methacrylate (PMMA) which is useful in implementation of the on-line control of polymerization process.