中医药数据隐结构分析平台---孔明灯网络版＊

隐结构分析是由香港科技大学张连文教授为研究中医药而首创的方法。该方法基于显变量的概率同现构建隐结构模型（又称隐树模型）,能够较好地模拟中医诊病推理的过程,得到了同行的认可。但是,与方法同步发行的隐结构模型分析软件（孔明灯）为单机版本,随着待分析数据量的增大,存在模型学习时间过长、占用计算机资源过多等问题。因此,为解决该方面不足,我们基于单机版的开放资源搭建了隐结构分析的网络平台版本,网址为http://ai.hactcm.edu.cn。该网络版本利用网络云资源,动态调整所需计算资源,有三个计算节点能够并发支持200个用户同时使用,并实现7*24小时全天候计算,在模型学习的效率上较单机版提升5倍以上。同时网络平台还具有业务日志、登录日志、监控管理等多个附加功能,较大地方便了用户和管理人员,为推广和使用隐结构研究中医药起到积极的推动作用。

The Web Version of Lantern: A Online Analysis Platform for TCM based on Latent Tree

Latent structure analysis is a method pioneered by professor Nevin Zhang of the Hong Kong university of science and technology that for researching Traditional Chinese Medicine. Based on explicit variable probabilistic co-occurrence to construct the latent structure model (also known as latent tree model), this method can simulate the reasoning process of diagnosis of traditional Chinese medicine, and has been recognized by many peers. However, the latent structure model analysis software (Lantern) released simultaneously with the method is a stand-alone version. With the increase of the amount of data to be analyzed, problems such as too long model learning time and too much use of computer resources are existing. Therefore, in order to solve this problem, we built an online platform version of the latent structure model, which is http://ai.hactcm.edu.cn. The network version makes use of network cloud resources to dynamically adjust the required computing resources. Three computing nodes can support 200 users simultaneously, and realize 7*24 hours round-the-clock computing, which improves the efficiency of model learning by more than 5 times compared with the stand-alone version. At the same time, the network platform also has many additional functions, such as business log, monitoring and management, etc., which facilitates users and managers in large places and plays an active role in promoting and using hidden structure research on Traditional Chinese Medicine.