A Two-Stage Learning Method for Response Prediction

摘要: Social networks have become a popular and powerful communication platform as the mobile technology evolves. To evaluate influence of message on social network, response prediction is crucial in modeling propagation interaction among users. predict whether new will receive responses, we propose two-stage learning method using set features derived from messages, users theirresponding behaviors. This first clusters then learns different models respectively. The central argument for this strategy that classifiers trained separately clustered data sets can focus particular types data, reduce effects noise, consequently an overall higher predictive performance than single classifier entire set. We tested proposed learner Plurk, compared it withother classifiers. experimental results show outperformed gradient boosting decision tree learner, logistic function support vector machine not only accuracy, but also efficiency.