摘要
近年来随着"IP"热潮兴起,网络文学市场发展迅速,逐渐成为文化娱乐行业投资热点.本文将机器学习方法引入到小说排行预测方面,通过网络爬虫获取网络小说信息并提取了影响排行的特征,提出了基于BP神经网络模型进行小说排行预测.针对训练数据的不均衡,本文采用ROC和AUC作为预测评价指标;实验结果表明,基于BP神经网络的网络小说排行预测的准确率较高,相比传统的文学定性分析方法,机器学习预测方法可解释性和应用性更高.
With the rise of the"IP"boom in recent years,the online literature market is developing rapidly and has gradually become a hotspot for investment in the cultural and entertainment industry.This paper introduces the machine learning method to the novel ranking prediction,the characteristics of the influence rankings are extracted from the network novel(also called online novel)information collected by the web crawler,a BP neural network model is developed to predict the range of the online novel rankings.As the training dada is unbalanced,ROC and AUC are selected as the evaluation indicators of prediction.The experimental results show that the accuracy of online novel ranking prediction based on BP neural network is more accurate.Compared with traditional literary qualitative analysis method,machine learning approach is interpretable and applicable in the online novel ranking prediction.
作者
龙彬
胡思才
郭峻铭
李旭伟
LONG Bin;HU Si-Cai;GUO Jun-Ming;LI Xu-Wei(College of Computer Science,Sichuan University,Chengdu 610065,China;Unit 78179,PLA,Chengdu 611130,China;Unit 61920,PLA,Chengdu 610505,China)
出处
《四川大学学报(自然科学版)》
CAS
CSCD
北大核心
2019年第1期50-56,共7页
Journal of Sichuan University(Natural Science Edition)
基金
国家自然科学基金(61173099)