Research on Video Ratings Prediction and Survival Analysis
- Resource Type
- Conference
- Authors
- Fulian, Yin; Yueqi, Jiang; Pei, Su; Ge, Su
- Source
- 2017 International Conference on Computer Systems, Electronics and Control (ICCSEC) Computer Systems, Electronics and Control (ICCSEC), 2017 International Conference on. :381-384 Dec, 2017
- Subject
- Communication, Networking and Broadcast Technologies
Components, Circuits, Devices and Systems
Computing and Processing
Robotics and Control Systems
Autoregressive processes
Correlation
Predictive models
Analytical models
Biological neural networks
Ratings prediction
Survival analysis
BP neural network
Autoregressive moving average
Cox proportional hazard model
- Language
Aiming at the poor accuracy of the existing Internet video ratings prediction, this paper proposes a self-correlation prediction based on the view amount and a survival analysis scheme based on the video component factors. The self-correlation prediction scheme uses BP neural network and autoregressive moving average (ARMA) to predict the short video, and a multi-direction hybrid prediction method based on the ladder data to predict the long video. The experimental results show that the prediction accuracy of the above methods can reach up to 90%. The survival analysis scheme adopts the COX proportional hazard model (COX model) to analyze the correlation between the component factors of the variety shows and their ratings, then predicts the ratings according to these factors. The experimental results show that the concordance index (C-index) can be up to 0.8.