捕鱼算法优化核极限学习机的微博热点话题预测

Hot topic prediction of micro-blog based on kernel extreme learning machine and fishing algorithm

微博热点话题预测对网络舆情控制与管理具有重要意义,针对微博网络热点话题的随机性、非线性以及核极限学习的隐层权值和隐层阈值优化难题,提出一种捕鱼算法优化在核极限学习的微博热点话题预测模型。首先将微博网络热点话题历史样本划分训练样本和测试样本集,然后采用在核极限学习对微博热点话题训练样本进行学习与建模,并采用捕鱼算法优化在线极限学习的隐层权值和隐层阈值,最后采用微博热点话题测试样本对其性能进行测试。实验结果表明,本文模型可以描述微博热点话题的发展趋势,提高了网络热点话题的预测精度,而且性能优于其它网络热点话题预测模型。

hot topic prediction of micro-blog has a important meaning to control and manage public opinion in net-work, according to the micro-blog network hot randomness, nonlinearity and online extreme learning hidden layer weights and hidden layer threshold optimization problems, this paper put forward a hot topic prediction model of micro-blog based on online extreme learning machine by fishing algorithm in this paper. Firstly, historical samples of hot topic for micro-blog are divided into training samples and testing samples set, and secondly, extreme learning machine is used to learn training samples of micro-blog hot topic, finally, test samples of micro-blog hot topic are used to test the performance. The experimental results show that the proposed model can describe the development trend of micro Bo hot topics, improves the prediction accuracy of hot topic, and the performance is better than that of other network hot topic prediction model.