Emotional e-mail classification is one of the important issues in the service oriented organizations. E-mails are served in a first come first serve basis. Few e-mails express the unfair treatment or dissatisfaction of service. It is essential to serve such e-mails with a high priority. In this paper an attempt is made to identify such mails which express the strong emotions of the customers/stakeholders. This system classifies the e-mails in to three categories via positive, negative and other mails. An adaptive machine learning algorithm that uses combined SVD and KNN methods is developed to solve the problem of emotional e-mail classification. Also an emotional dictionary is used as a central component of this system that serves various emotional words and phrases for classification. The system also adaptive in nature and adapts various new words and phrases that explicates the emotion.