基于数据预处理深度置信网络的药物与非药物分类（英文）

Classification of Drug and Non-Drug Using Deep Belief Network Based on Data Preprocess

制药工业的一个主要趋势是整合传统意义上被认为早期阶段药物发现的分子描述.为了更好的将药物和非药物分类,本文提出了基于深度信念网络(DBN)的分类模型.首先,对分子特征进行预处理以保证有价值的信息得到保留,其次,该模型将DBN和反向传播(BP)分类器结合去对药物/非药物进行检测和分类.DBN由几个受限玻尔兹曼机(RBM)层组成,当特征向量转移到下一层时这些RBM层尽可能多的保留具有重要的影响的信息.BP层训练的最后一个RBM层生成特征分类.结果表明,该方法是提取高层次特征的药物和非药物分类任务中一种成功的方法,分类精度高达85.3%,高于传统的支持向量机和神经网络方法.同时,预处理对分子特征的提取更为有效,从而在一定程度上提高了分类的准确性.

One of the key trends in the pharmaceutical industry has been the integration of what have traditionally been considered as molecular descriptions of the early phases of drug discovery. In order to better classify drug and non-drug,a classified model based on deep belief network（DBN） is proposed in this paper. Firstly, the preprocessing of molecular features to guarantee the valuable information is retained. Secondly, the model is a hybrid of DBN and Back Propagation（BP） classifier to detect and classify drug/non-drug. The DBN builds consists of several restricted boltzmann machine（RBM） layers, which maintain as much information with important influence as possible when the feature vectors are transferred to next layer. The BP layer is trained to classify the features generated by the last RBM layer. The results showed that the method is a successful approach in the high-dimensional-feature for drug and non-drug classification task as the useful high-level features are extracted. The classified accuracy is up to 85.3% which is higher than the traditional methods such as support vector machine（SVM） and traditional neural network（NN）. Meanwhile, the pre-process is more effective extract for molecular features so that the accuracy of classification has been improved to certain extent.