基于神经网络的刑事案件量刑决策系统  

Sentence Penal Measure System Based on Neural Network

作　　者：邹晓玫 修春波[2] 

机构地区：[1]天津商学院法学系,天津300134 [2]天津工业大学计算机技术与自动化学院,天津300160

出　　处：《微计算机信息》2008年第3期264-265,共2页Control & Automation

基　　金：天津商学院青年科研培育基金(No.051116);国家自然科学基金(No.10402003)

摘　　要：为了给刑事案件量刑提供一个客观公正的参考,提出了一种基于神经网络的量刑决策系统。根据案件的具体细节情况,确定案件各种量刑情节的有无,并将其作为分类和决策的特征量,确定刑罚的种类以及刑期的长短。为提高网络的决策性能,针对梯度学习存在的局部极小和假饱和等现象,提出了基于混沌退火的学习算法,进一步提高了网络的决策能力。大量的仿真实验表明,该决策网络能够给出一个客观、公正的量刑结果。To provide an objective and equitable reference result,sentence penal measure system based on neural network was proposed.Sentencing circumstances could be determined according to the case's detailed characters,which would be used as characteristic parameter to determine the penalty category and the term of imprisonment.Combining chaos simulated annealing with gradient search a novel learning method was proposed to improve the network's decision performances.Numbers of simulation experiments prove that the ...

关 键 词：神经网络 量刑 混沌 决策系统 

分 类 号：TP319[自动化与计算机技术—计算机软件与理论]