基于AHP-LSTM的校园网络沉迷判定方法研究

AHP-LSTM Based Evaluation Methods on Addiction Behaviors in Campus Networks

随着电信技术的发展,校园网络在高等教育中发挥着重要作用。然而,目前大学生沉迷网络现象日益严重。因此,如何防止大学生沉迷网络是大学教育和管理的一项重要任务。针对该问题,设计出一种包含数据获取、沉迷判定和网络管控的高校校园网络沉迷系统。针对目前存在的3种常见网络沉迷行为,即网络游戏沉迷、在线视频沉迷和网站论坛沉迷,提出了基于网络沉迷指数的评价体系和量化标准,并选取了4个影响网络沉迷指数计算的关键指标,即总时长、时间占比、频率、行为丰富度,制定出了沉迷指数的计算和预测方案,即采用层次分析法对网络沉迷行为及关键指标进行优化建模,并基于长短期记忆网络(LSTM)对沉迷指数进行预测和验证。最后通过计算机仿真得出LSTM预测模型的均方根误差为3.1386,相比其他预测方法该值最低,证实了所提网络沉迷判定和预测方法具有较高的可行性。

With the development of telecommunication technology,campus network plays an important role in higher education.However,the phenomenon of internet addiction among college students is becoming increasingly serious at present.Therefore,how to keep university students away from being addicted into the network is an important task in university education and management.In response to this problem,an anti-addiction system for college students has been proposed,which includes data obtaining,addiction behavior evaluation and network management.Focusing on three common types of internet addiction including online games,online video and website forums,an evaluation system and quantitative standard based on network addiction index is proposed.For the three common types of internet addiction behaviors currently identified such as addiction to online gaming,online video watching and website/forum browsing,an evaluation system and quantitative criteria based on the internet addiction index is proposed.Four key indicators affecting the calculation of the internet addiction index have been selected including total duration,proportion of time,frequency and behavior diversity.A calculation and prediction scheme for the addiction index is established,which involves the application of the analytic hierarchy process for the optimization modeling of internet addiction behaviors and key indicators,and the use of long short-term memory(LSTM)neural networks to predict and verify the addiction index.Finally,the root mean square error of LSTM prediction model is 3.1386 by computer simulation,suggesting that is the lowest compared with the other prediction methods,which proves that the network addiction judgment method is feasible.