ACADEMICIA: An International Multidisciplinary Research Journal

In artificial intelligence (AI), most of the application domain have the deficiency of not having their data labeled. Even some that are defined are not suitable for semi-supervised learning method. Although, unlabeled data are cheaply available in the public domain, it is not often used in artificial intelligence applications. Labeled data ensures that the AI support-application have clear algorithms for implementation. To get labeled instances, it is very difficult because experienced domain experts are required to label the unlabeled data patterns. Semi-supervised learning method addresses this problem and act as a middle-man between supervised and unsupervised learning. It is therefore an improved self-learning algorithm for AI implementations. This work addresses few techniques of enhanced semi-supervised learning (SSL) such as self-training and co-training. A new algorithm was developed that uses two approaches in a congested network environment that is being managed with an intelligent switch. The trained switch can control number of users on the network for effective data communication without interference and packet loss. This work could be beneficial to network managers, schools with large network infrastructures, to banks, to governments and to any other organization that deals or handles any topology of networks.