Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Short-Form Video Viewing Behavior Analysis and Multi-Step Viewing Time Prediction

Vu Thi Hai Yen, Duc V. Nguyen, Cao Anh Minh Huy, Truong Thu Huong

Abstract

Short-form videos have become one of the most popular user-generated content formats nowadays. Popular short-video platforms use a simple streaming approach that preloads one or more videos in the recommendation list in advance. However, this approach results in significant data wastage, as a large portion of the downloaded video data is not used due to the user's early skip behavior. To address this problem, the chunk-based preloading approach has been proposed, where videos are divided into chunks, and preloading is performed in a chunk-based manner to reduce data wastage. To optimize chunk-based preloading, it is important to understand the user's viewing behavior in short-form video streaming. In this paper, we conduct a measurement study to construct a user behavior dataset that contains users' viewing times of one hundred short videos of various categories. Using the dataset, we evaluate the performance of standard time-series forecasting algorithms for predicting user viewing time in short-form video streaming. Our evaluation results show that Auto-ARIMA generally achieves the lowest and most stable forecasting errors across most experimental settings. The remaining methods, including AR, LR, SVR, and DTR, tend to produce higher errors and exhibit lower stability in many cases. The dataset is made publicly available at https://nvduc.github.io/shortvideodataset.

Short-Form Video Viewing Behavior Analysis and Multi-Step Viewing Time Prediction

Abstract

Short-form videos have become one of the most popular user-generated content formats nowadays. Popular short-video platforms use a simple streaming approach that preloads one or more videos in the recommendation list in advance. However, this approach results in significant data wastage, as a large portion of the downloaded video data is not used due to the user's early skip behavior. To address this problem, the chunk-based preloading approach has been proposed, where videos are divided into chunks, and preloading is performed in a chunk-based manner to reduce data wastage. To optimize chunk-based preloading, it is important to understand the user's viewing behavior in short-form video streaming. In this paper, we conduct a measurement study to construct a user behavior dataset that contains users' viewing times of one hundred short videos of various categories. Using the dataset, we evaluate the performance of standard time-series forecasting algorithms for predicting user viewing time in short-form video streaming. Our evaluation results show that Auto-ARIMA generally achieves the lowest and most stable forecasting errors across most experimental settings. The remaining methods, including AR, LR, SVR, and DTR, tend to produce higher errors and exhibit lower stability in many cases. The dataset is made publicly available at https://nvduc.github.io/shortvideodataset.
Paper Structure (15 sections, 10 equations, 5 figures)

This paper contains 15 sections, 10 equations, 5 figures.

Table of Contents

  1. Introduction
  2. Related Work
  3. User Viewing Time Data Collection
  4. Data Collection
  5. Data Analysis
  6. Viewing Time Prediction
  7. Problem Setup
  8. Prediction Models
  9. Linear Regression (LR)
  10. AutoRegression (AR)
  11. Auto-ARIMA
  12. Support Vector Regression (SVR)
  13. Decision Tree Regression (DTR)
  14. Evaluation Protocol and Metric
  15. Conclusion

Figures (5)

  • Figure 1: Spatial and temporal information values of short videos used in our experiment.
  • Figure 2: Screenshots of the custom-developed short video streaming application.
  • Figure 3: CDF of watch-time retention ratio aggregated over all viewing sessions from 50 users
  • Figure 4: CDF of watch-time retention ratio across different video categories
  • Figure 5: Comparison of the RMSE of different methods for $N=10$ to $N=80$.