Social Media Prediction Challenge

News

• [06-16-2019]: Challenge submission already closed.
• [06-11-2019]: Submission already opened, please refer to the link after logged in
• [06-08-2019]: TEST SET already released now!
• [04-16-2019]: We have released the training data.
• [03-25-2019]: We are hosting Social Media Prediction Challenge (ACM Multimedia 2019).

Overview

SMP Challenge (Social Media Prediction Challenge) is a annual prediction challenge to seek excellent research teams for prediction and discover novel and emerging challenges based on numerous resources of social multimedia. Making predictions via social multimedia (photos, video or news) is not only helps us to make better strategic decisions for future, but also explore advanced predictive learning and analytics methods for various of problems and scenarios in several areas, such as multimedia advertising, recommendation system, user engagement, and trend analysis etc.

As a joint activity with the research teams from multiple organizations, we are holding on the SMP Challenge 2019 continuously. In this year, the open task of SMP Challenge is Temporal Popularity Prediction, which focused on predicting future clicks of new social media posts before they were posted in social feeds. The participated teams need to design new algorithm based on understanding and learning techniques, and automatically predict popularity (formulated by clicks or visits etc.) to achieve better performances.

SMPD2019 is a large social multimedia dataset, which contains over 486K records from 69K users. And each of social media posts has rich contextual information and annotations (e.g. image, text, temporal-spatial information, user profile, etc.).

Dataset Statistics

The SMPD2019 (Social Media Prediction Dataset) contain 486K social multimedia posts with web images, text and contextual data and rich annotations, which collected from Flickr (one of the largest photo-sharing platform) and corresponding anonymized photo-sharing records. SMPD2019 is a multi-faced data collection, which contains user profile, post category, customize tag, geography information, photo image, and photo metadata. For the prediction task, we split the data with time-order, resulting in train and test ratio is 2:1. The tables below show the statistics of the dataset.

Important Dates

Citation

If you intend to publish results that use the information and resources provided by this challenge, please include the following references:

@inproceedings{Wu2017DTCN,
  title={Sequential Prediction of Social Media Popularity with Deep Temporal Context Networks},
  author={Wu, Bo and Cheng, Wen-Huang and Zhang, Yongdong and Qiushi, Huang and Jintao, Li and Mei, Tao},
  booktitle={International Joint Conference on Artificial Intelligence (IJCAI)},
  year={2017},
  location = {Melbourne, Australia}}

@inproceedings{Wu2016TemporalPrediction,
  author = {Wu, Bo and Mei, Tao and Cheng, Wen-Huang and Zhang, Yongdong},
  title = {Unfolding Temporal Dynamics: Predicting Social Media Popularity Using Multi-scale Temporal Decomposition},
  booktitle = {Proceedings of the Thirtieth AAAI Conference on Artificial Intelligence (AAAI)}
  year = {2016},
  location = {Phoenix, Arizona}}

@inproceedings{Wu2016TimeMatters,
  author = {Wu, Bo and Cheng, Wen-Huang and Zhang, Yongdong and Mei, Tao},
  title = {Time Matters: Multi-scale Temporalization of Social Media Popularity},
  booktitle = {Proceedings of the 2016 ACM on Multimedia Conference (ACM MM)},
  year = {2016},
  location = {Amsterdam, The Netherlands}}
            

Our Teams

Bo Wu

Columbia University

bo.wu@columbia.edu

Wen-Huang Cheng

National Chiao Tung University

whcheng@nctu.edu.tw

Bei Liu

Microsoft Research Asia

bei.liu@microsoft.com

Jiebo Luo

University of Rochester

jluo@cs.rochester.edu

Zhaoyang Zeng

Sun Yat-sen University

zengzhy5@mail2.sysu.edu.cn

Peiye Liu

Beijing University of Posts and Telecommunications

liupeiye@bupt.edu.cn