基于基因限制信息的特征权值优化研究

Study of Feature Weight Optimization Based on Gene Restrictive Information

针对自组织映射算法在衡量神经元与输入数据之间相似度时将所有特征视为权值相同,导致无法在基因半监督聚类中利用限制信息的某些重要特征的不足,提出了一种新的基于基因限制信息的特征权值优化算法。算法根据限制信息优化特征的权值,将那些能够有效区分限制信息的特征赋予较大的权值。为防止在特征权值的优化中引入偏置或对非限制数据划分能力减弱,算法利用非限制信息调整特征权值。实验结果表明,算法提高了基因聚类的准确率。

SOM sets the same weight to all features and it is impossible to take advantage of some essential feature in gene semi-supervised clustering. A novel feature weight optimization algorithm based on restrictive information is proposed. The restrictive information is used to optimize the feature＇s weight and give more attention to the data that effectively distin guish restrictive information. To avoid the bias or the weak ability to cluster nonrestrictive data, the non-restrictive data is also used to adjust the feature＇s weight. Experiment results show the accuracy of gene clustering is improved.