基于RS-BP神经网络的政务微信公众号信息质量评价模型研究

Information Quality Evaluation Model Research of Government WeChat Public Account Based on RS-BP Neural Networks

【目的/意义】政务微信公众号是政府沟通民众的重要通道,对政务微信公众号信息质量进行评价研究,有助于政府树立威信形象、提升政府的公信力。【方法/过程】基于"5W传播"模型从信息生产力、信息内容、信息表现、用户、信息影响力五个维度初步获取39项评价指标,利用粗糙集理论(RS)约简为11项核心指标并进行综合评价,将评价结果作为输入数据对BP神经网络进行仿真模拟训练,训练成功后生成政务微信公众号RS-BP神经网络信息质量评价模型,最后选取江西省4个不同类型5个典型政务微信公众号进行实证研究。【结果/结论】实证研究表明,本模型在处理政务微信公众号信息质量评价这种非线性问题时具有一定的适用性,可为政务信息质量评价和改善提供参考。【创新/局限】研究所生成的政务微信公众号RS-BP神经网络信息质量评价模型具有一定的实用价值,但调查对象种类不够广泛,训练数据不足,模型成熟度仍需进一步完善。

【Purpose/significance】The government WeChat public account is an important channel for the government to communicate with the public. The evaluation of the information quality of the government WeChat public account helps the government to establish a prestige image and enhances the government’s credibility.【Method/process】Based on the"5W propagation"model, 39 evaluation indicators were initially obtained from the five dimensions of information productivity, information content, information performance,users and information influence. The Rough Set theory(RS) reduction was used as 11 core indicators and comprehensive evaluation was carried out. The evaluation results are used as input data to simulate and train the BP neural networks. After training successfully,the RS-BP neural networks information quality evaluation model of the government WeChat public accounts is generated, which is applied to the empirical analysis of 4 different types of typical government WeChat public accounts in Jiangxi.【Result/conclusion】The empirical research shows that this model has certain applicability in dealing with the non-linear problem of government WeChat public accounts information quality evaluation, which can provide reference for the evaluation and improvement of government information quality.【Innovation/limitation】The RS-BP neural networks information quality evaluation model of the government WeChat public accounts has practical value, but the types of survey objects are not extensive enough, and training data is insufficient. The maturity of the model needs further improvement.