Viral marketing is a form of online word-of-mouth (WOM) communication in which individuals are encouraged to pass on promotional messages through social websites. With the growing popularity of online social websites, viral marketing has increasingly garnered attention of marketers and marketing researchers alike. The two most important WOM attributes highlighted in the extant literature are volume and valence. This thesis looked into the cause, development and outcome of WOM marketing and provided computational models for forecasting the development of WOM volume and valence of viral marketing in social websites. With the data extracted from large-scale web-crawling activities, through a series of computer simulation experiments comparable to social websites, the author developed models to predict WOM volume and valence in viral marketing. The model for predicting WOM volume in viral marketing used theories of network topologies. The model for predicting WOM valence in viral marketing used an artificial neural network model. The author discussed the insights from the findings and suggested viral marketing strategies to optimize the performance of WOM volume and valence in social websites. A key contribution of this thesis is the new approaches of modeling and data collection for WOM volume and valance forecasting in viral marketing.