The Diffusion Of Semiconductor In China Based On Extended Mulitple Generation Innocaiton Diffusion Model

Semiconductor industry is capital,and has a high technological content and a biginvestment risk.With the globalization of economy development, market competitionbecomes increasingly fierce, in order to maintain the market competitive advantage,thehigh technology products manufacturer constantly increase spending on research, andlaunch multiple generation technology innovation products. This speeds up productsrenewal speed, and shortens product lifecycle. In addition, Compared with the previousgeneration product,the newer have better performance.The new products do not onlyopen up new market,but also divide up the old generation product's market share.Itmakes the diffusion of every generation product be impacted by the older and the newer.Therefore,it is great significance to research the diffusion law of multiple generationinnovation product, and this is helpful for enterprises to grasp the diffusion trend ofeach generation product and make corresponding marketing strategy.In the innovation diffusion model research, the classic innovation diffusion model,such as the Bass model, Mansfield model,and so on, they mainly research a singleproduct diffusion process. In1987, Norton and Bass based on the classic innovationdiffusion model, and firstly put forward a multiple generation innovation productdiffusion model, Norton-Bass model. Due to the condition of Norton-Bass model is veryhard, and too idealistic, therefore, the later scholars based on the Norton-Bass model,and relaxed the model assumption gradually.They put forward many more generationsinnovation diffusion model by joining some new innovation diffusion influence factorsin the model. These models include:①introducing repeat purchase of multiplegeneration innovation diffusion model;②introducing marketing variables,such asprice,of innovation diffusion model;③introducion leapfrogging of generationsinnovation diffusion model;④introducing market growth rate of innovation diffusionmodel for several generations. When these models were used to analysis multiplegeneration product innovation diffusion process in real life,all of them have their own advantages.but they just extended models from a certain aspects.In the real world,thediffusion process of innovative products are effected by various factors.Based on the Norton-Bass multiple generation innovation diffusion model, thispaper comprehensively considers the price, repeat purchase, market growth rates andseasonal factors which influence product innovation diffusion, and construct a newmultiple generation innovation diffusion model SMPRT.Through analysising andcomparisoning various parameter estimation methods, this paper finally decides to usegenetic algorithm as the model's parameters estimation method, and research on thesemiconductor manufacturing technology.The model abtains satisfactory results.Thepaper compares the research results of SMPRT model with that of Norton-Bass modeland Islam-Meade model. The results show that the SMPRT model is better thanNorton-Bass model and Islam-Meade model in goodness-of-fit and predictionaccuracy.So the SMPRT model to study the multiple generation product innovationdiffusion is a better choice. In addition, this paper compares genetic algorithm(GA) withnonlinear least squares method(NLS) and maximum likelihood estimationmethod(MLE).And the results find that the estimation effect of genetic algorithm ismore satisfactory. Finally, based on the genetic algorithm,and using SMPRT innovationdiffusion model,this paper forecasts the diffusion trend of the wafer.