Browsing by Author "Ting, Zih-Yun"

Now showing 1 - 2 of 2

A Similarity-Based Learning Algorithm Using Distance Transformation
(2015-06-01)
Hu, Yuh-Jyh
;
Yu, Min-Che
;
Wang, Hsiang-An
;
Ting, Zih-Yun
;
交大工研院聯合研發中心
Numerous theories and algorithms have been developed to solve vectorial data learning problems by searching for the hypothesis that best fits the observed training sample. However, many real-world applications involve samples that are not described as feature vectors, but as (dis) similarity data. Converting vectorial data into (dis) similarity data is more easily performed than converting (dis) similarity data into vectorial data. This study proposes a stochastic iterative distance transformation model for similarity-based learning. The proposed model can be used to identify a clear class boundary in data by modifying the (dis) similarities between examples. The experimental results indicate that the performance of the proposed method is comparable with those of various vector-based and proximity-based learning algorithms.
A Two-Stage Learning Method for Response Prediction
(2015)
Chen, Kuan-Hsi
;
Ting, Zih-Yun
;
Shen, Jia-Ying
;
Hu, Yuh-Jyh
;
Liang, Tyne
Social networks have become a popular and powerful communication platform as the mobile technology evolves. To evaluate the influence of a message on a social network, response prediction is crucial in modeling the message propagation and interaction among users. To predict whether a new message will receive responses, we propose a two-stage learning method using a new set of features derived from the messages, the users and their responding behaviors. This method first clusters the messages, and then learns the different prediction models from the clusters respectively. The central argument for this two-stage strategy is that the classifiers trained separately from the clustered data sets can focus on particular types of data, reduce the effects of noise, and consequently have an overall higher predictive performance than a single classifier trained from the entire data set. We tested the proposed two-stage learner on Plurk, and compared it with other classifiers. The experimental results show that the two-stage learner outperformed the gradient boosting decision tree learner, the logistic function learner, and the support vector machine for not only the predictive accuracy, but also for the efficiency.