Mining Interest Diffusion in Online Activity Data Streams

Published: 01 Jan 2024, Last Modified: 06 Aug 2024WWW (Companion Volume) 2024EveryoneRevisionsBibTeXCC BY-SA 4.0
Abstract: Modeling and forecasting such data is difficult because online activity data is high-dimensional and composed of multiple time-varying dynamics such as trends, seasonality, and diffusion of interest. In this paper, we propose D-Tracker, designed to capture latent dynamics in online activity data streams and forecast future values. Our proposed method has the following properties: (a) Interpretable: it uses interpretable differential equations to model the latent dynamics in online activity data, which enables us to capture trends and interest diffusion among locations; (b) Automatic: it determines the number of latent dynamics and the number of seasonal patterns fully automatically; (c) Scalable: it incrementally and adaptively detects shifting points of patterns for a semi-infinite collection of tensor streams. (c)Scalable : the computation time of D-Tracker is independent of the time series length. Experiments using web search volume data obtained from GoogleTrends show that the proposed method can achieve higher forecasting accuracy in less computation time than existing methods while extracting the patterns of interest diffusion among locations.
Loading