Modeling Customer Satisfaction Considering Kansei Requirements And Its Application On Product Customization

As customer's attitude to consumption changes and function homogenization becomes more serious,product hardware design has more and more influence on buying decision making,in which customer's characteristics and emotional factors will play more important role.For corporations that aim for profiting,it's not doubt important to find a method to predict satisfaction based on product features and remark data and predict customer's willingness to buy a product and recommend optimal product accurately.In order to precisely predict satisfaction and fulfill optimal product recommendation,meanwhile providing advises for product design for corporations,the thesis proposed a novel approach to model satisfaction and accomplish product configuration avoiding shortcomings of tradition methods.Firstly,kansei engineering was applied to modeling for considering emotional requirements and customers were clustered based on emotional factors.Satisfaction appraise were categorized to multiple kansei dimensions.Due to the perception consistency emotionally,normal distribution influence can be erased.Secondly,when modeling satisfaction,random forest and neural network were employed respectively to select features and calculate kansei satisfaction.Finally,association rules mining were used to figure out basic design principles need to be followed and conflicted combinations need to be avoided.Bidirectional association rules-constrained genetic algorithm(BAR-GA)is proposed to limit freedom of configuration,which makes configuration result in the range of controlling and meanwhile can provide guidance for companies.Through comparing prediction error and recommend sample votes of novel approach and tradition one,the efficiency and accuracy of presented approach were proved,which can provide suggestions with guiding meaning for corporations to predict satisfaction,launch customer survey and new product design.