Abstract:Social media has become one of the most credible sources for delivering messages, breaking news, as well as events. Predicting the future dynamics of an event at a very early stage is significantly valuable, e.g, helping company anticipate marketing trends before the event becomes mature. However, this prediction is non-trivial because a) social events always stay with "noise" under the same topic and b) the information obtained at its early stage is too sparse and limited to support an accurate prediction. In order to overcome these two problems, in this paper, we design an event early embedding model (EEEM) that can 1) extract social events from noise, 2) find the previous similar events, and 3) predict future dynamics of a new event with very limited information. Specifically, a denoising approach is derived from the knowledge of signal analysis to eliminate social noise and extract events. Moreover, we propose a novel predicting scheme based on locally linear embedding algorithm to construct the volume of a new event from its k nearest neighbors. Compared to previous work only fitting the historical volume dynamics to make a prediction, our predictive model is based on both the volume information and content information of events. Extensive experiments conducted on a large-scale dataset of Twitter data demonstrate the capacity of our model on extract events and the promising performance of prediction by considering both volume information as well as content information. Compared with predicting with only the content or the volume feature, we find the best performance of considering they both with our proposed fusion method.

E-Embed: A time series visualization framework based on earth mover's distance.

Using Visualization to Improve Clustering Analysis on Heterogeneous Information Network.

Representation and analysis of time-series data via deep embedding and visual exploration

Visual Analytics of Economic Features for Multivariate Spatio-Temporal GDP Data

EmbeddingVis: A Visual Analytics Approach to Comparative Network Embedding Inspection

Embedding and Predicting the Event at Early Stage

EmbeddingTree: Hierarchical Exploration of Entity Features in Embedding

Statistical embedding: Beyond principal components

Transforming Multidimensional Time Series into Interpretable Event Sequences for Advanced Data Mining

I-SEA: Importance Sampling and Expected Alignment-Based Deep Distance Metric Learning for Time Series Analysis and Embedding

A Shapelet-based Framework for Unsupervised Multivariate Time Series Representation Learning

Clustering and Recognition of Spatiotemporal Features through Interpretable Embedding of Sequence to Sequence Recurrent Neural Networks

A Parallel Framework for Streaming Dimensionality Reduction

WizMap: Scalable Interactive Visualization for Exploring Large Machine Learning Embeddings

An integrated image visibility graph and topological data analysis for extracting time series features

Forecasting high-dimensional dynamics exploiting suboptimal embeddings

Empowering Interdisciplinary Insights with Dynamic Graph Embedding Trajectories

Casting Multiple Shadows: High-Dimensional Interactive Data Visualisation with Tours and Embeddings

Embedding Geographic Information for Anomalous Trajectory Detection.

Latent Space Cartography: Visual Analysis of Vector Space Embeddings

Emblaze: Illuminating Machine Learning Representations through Interactive Comparison of Embedding Spaces