Effectiveness of Action Prediction Method for a User using Inductive Learning with N-Gram

J. Xu, K. Araki, and K. Tochinai (Japan)


Inductive Learning, N-gram, Action prediction, Learning Room, Caring system, Adaptability


This paper describes a method for action prediction of a user. When we build a care system with learning function like a learning room, a statistical approach or an analytical approach can be considered. Statistical approaches are not liable to produce reliable result unless a huge prepared database is available. The analytical approaches are necessary to give the prepared rules adapted to the user and the adaptability of this method is low. Aiming at solution of such problems, we have proposed a method to predict the action of a user using Inductive Learning with N-gram. The system based on this method is able to acquire needed rules from comparative few data history automatically using Inductive Learning. The rules express a user's taste and custom. Therefore the system is able to adapt dynamically to the users by it’s own learning ability. The rate of the average correct prediction was 60.1[%] on the experiment. The user must proofread the erroneous conversion in the prediction results. However, the erroneous conversion decreases since the system based on this method is able to adapt dynamically to various users. This paper shows the evaluation results of the action prediction in our proposed method.

Important Links:

Go Back