Title: A similarity based learning framework for interim analysis of outcome prediction of acupuncture for neck pain

Authors: Gang Zhang; Zhaohui Liang; Jian Yin; Wenbin Fu; Guo-Zheng Li

Addresses: School of Information Science, SUN YAT-SEN University, Guangzhou 510000, China ' Hospital of Guangzhou University of Chinese Medicine, Guangzhou 510120, China ' School of Information Science, SUN YAT-SEN University, Guangzhou 510275, China ' Hospital of Guangzhou University of Chinese Medicine, Guangzhou 510120, China ' Guangdong Provincial Academy of Chinese Medical Sciences, Guangzhou 510120, China; Department of Control Science and Engineering, Tongji University, Shanghai 201804, China

Abstract: Chronic neck pain is a common morbid disorder in modern society. Acupuncture has been administered for treating chronic pain as an alternative therapy for a long time, with its effectiveness supported by the latest clinical evidence. However, the potential effective difference in different syndrome types is questioned due to the limits of sample size and statistical methods. We applied machine learning methods in an attempt to solve this problem. Through a multi-objective sorting of subjective measurements, outstanding samples are selected to form the base of our kernel-oriented model. With calculation of similarities between the concerned sample and base samples, we are able to make full use of information contained in the known samples, which is especially effective in the case of a small sample set. To tackle the parameters selection problem in similarity learning, we propose an ensemble version of slightly different parameter setting to obtain stronger learning. The experimental result on a real data set shows that compared to some previous well-known methods, the proposed algorithm is capable of discovering the underlying difference among different syndrome types and is feasible for predicting the effective tendency in clinical trials of large samples.

Keywords: interim analysis; kernel as similarity; decision tree; NDS; non-dominated sort; acupuncture; similarity based learning; outcome prediction; chronic neck pain; chronic pain; alternative therapy; machine learning; modelling; parameter selection; bioinformatics.

DOI: 10.1504/IJDMB.2013.056643

International Journal of Data Mining and Bioinformatics, 2013 Vol.8 No.4, pp.381 - 395

Received: 02 Mar 2012
Accepted: 02 Mar 2012

Published online: 20 Oct 2014 *

Full-text access for editors Access for subscribers Purchase this article Comment on this article