摘要
推荐系统已经应用到各行各业中,新闻推荐也应运而生。用户在阅读新闻时一般是时间序列的形式,然而,传统的新闻推荐算法很少考虑用户浏览行为的时间序列特征。因此,它并不能有效地预测用户阅读的下一篇新闻。针对这一问题,将隐马尔可夫模型融入到新闻推荐算法中,根据用户的阅读轨迹,找到用户下一时刻阅读概率最高的新闻。在此基础上,加入状态驻留时间元素,将隐马尔可夫模型的五元素扩展为六元素,以此来提高推荐准确度。为证明解决方案的有效性,与其他的新闻推荐算法进行了对比,结果显示论文算法在F1指标上约提高了14%。
Recommendation system has been applied to all walks of life,and news recommendation has emerged as the times require.Users usually read news in the form of time sequence.However,traditional news recommendation algorithms seldom consider the time sequence characteristics of users browsing behavior.Therefore,it can not effectively predict the next news that users read.To solve this problem,the paper incorporates hidden Markov model into news recommendation algorithm.According to the user's reading trajectory,it can find the news with the highest reading probability at the next moment.On this basis,it adds the state dwell time element to extend the five elements of hidden Markov model to six elements to improve the accuracy of recommendation.To prove the effectiveness of our solution,it compares with other news recommendation algorithms and finds that our algorithm increased by about 14%on the F1.
作者
张丹
周从华
ZHANG Dan;ZHOU Conghua(School of Computer Science and Telecommunication Engineering,Jiangsu University,Zhenjiang 212013)
出处
《计算机与数字工程》
2020年第10期2332-2337,共6页
Computer & Digital Engineering
关键词
新闻推荐
时间序列
隐马尔可夫模型
阅读轨迹
驻留时间
news recommendation
time sequence
hidden Markov model
reading trajectory
dwell time
