Performance Analysis of Random Forest and K-nearest Neighbour for Prediction of Video Transcoding Time

Currently, the distribution of video content to numerous devices via different communication networks cannot be over-emphasized. This distribution must be adjusted to the resolution, size of the screen, bandwidth availability, and the decoding power of the final user gadgets. This paper studied the effectiveness of the Random Forest and K-Nearest Neighbour for predicting the video transcoding time. Video transcoding executes several tasks which include format conversions and bit rate readjustment, to convert one video stream compression to another. Transcoding helps diverse media with different devices formats and capabilities to exchange video content on heterogeneous network platforms such as the Internet. The two algorithmic models predict the time of transcoding of video segments for the codecs using the transcoding parameters and content complexity. The performance of the proposed random forest and KNN is evaluated using online video characteristics and transcoding. Experimental results showed Random forest outperforms the KNN in all the metrics utilized. Furthermore, the Random Forest is better in terms of performance when compared with other approaches available in the literature with a very low MAE of 0.809 and had a competitive performance in <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$R^{2}$</tex>.