Research On Text Classification Of Active Learning And Its Application

Since the birth of artificial intelligence,related theory and technology matures,application fields are also expanding.As an important tool to realize the artificial intelligence,machine learning algorithms focus on how to make computers simulate human learning behaviors.Active learning must be on research until that the computer are able to do some complex work which would be done usually by human intelligence.This makes it possible to let the machines ask human for help in right time during the process of the task,then they can turn it into their own experience and use it in another task in the future.This article stars from the concept of active learning firstly,then introducing the related theory of the basic law and the development of active learning.Elaborating several classical active learning algorithms so far according to the logical relationship,which has formed a whole view about text active learning.Raising a question about the impact of using active learning to train models with an incomplete data sets,and then proposing an algorithm framework that can overcome the question mentioned.We put forward the IGAKME clustering algorithm,which can be implemented with distributed computing programming models,as the choosing method.On the basis of the above,we put forward an active learning classification method named SVMAL-IGAKME,which is based on the self-training model and C-SVC model.And then we design a series of experiments to prove that the algorithm can train an effective model of text classification using default training data,which is make up of real text data.The characteristics of SVMAL-IGAKME can be applied to some short text classification problem with frequent refresh rate,that is full of practical significance.