面向教学资源的均值惩罚随机森林非平稳时序预测方法

Mean Penalty Random Forest Nonstationary Time Series Prediction Method for Teaching Resources

教学资源数据信息因多种因素影响,往往随时间推移呈现出周期变化和随机变动等特性,几种变化特性组合或叠加造成其时间序列的非平稳性.这些数据蕴含着大量的多源信息,如老师学生交互数据、学生请求资源数据等,合理地对面向教学资源非平稳时序进行预测可以有效地促进在线教育平台提高用户使用体验感等.本文提出一种均值惩罚随机森林非平稳时序预测方法(PMP-RF),对通过了非平稳性检测的教学资源时序数据进行均值惩罚处理,采用随机森林模型对均值惩罚后的非平稳时序进行预测,从而得到预测值.对教学资源请求数据的预测结果表明,PMP-RF比传统的时间序列预测方法以及神经网络预测模型的精度更高,可以很好地应用在噪声较少的非平稳时序预测中.

Due to the influence of many factors,the data information of teaching resources often exhibits characteristics such as periodic changes and random changes over time.The combination or superposition of several change characteristics causes the nonstationarity of its time series.These data contain a large amount of multi-source information,Such as teacher-student interaction data,student request resource data,etc.,reasonable prediction of nonstationary time series for teaching resources can effectively promote online education platforms to improve user experience,etc.This paper proposes a mean penalty random forest nonstationary time series prediction method(PMP-RF),the average penalty is performed on the time series data of teaching resources that have passed the nonstationary detection,and the random forest model is used to predict the nonstationary time series after the average penalty,thereby obtaining the predicted value.Prediction of teaching resource request data The results showthat PMP-RF has higher accuracy than traditional time series prediction methods and neural network prediction models,and can be well applied to nonstationary time series prediction with less noise.