【数据科学赛】Global Knowledge Tracing Challenge @AAAI2023 #时间序列预测 #知识追踪

CompHub 实时聚合多平台的数据类(Kaggle、天池…)和OJ类(Leetcode、牛客…）比赛。本账号会推送最新的比赛消息，欢迎关注！

更多比赛信息见 CompHub主页 或 点击文末阅读原文

---

以下内容摘录自比赛主页

Part1赛题介绍

题目

Global Knowledge Tracing Challenge @AAAI2023(详细介绍)

举办平台

CondaLab

主办方

背景

在本次比赛中，我们希望呼吁全世界的研究人员和从业者通过具有丰富辅助信息的知识追踪方法来研究提高学生评估表现的机会。知识追踪（Knowledge Tracing, KT）是利用学生的历史学习交互数据，对他们的知识掌握情况进行建模，以预测他们未来的交互表现，如下图所示。当与高质量的学习材料和指导相结合时，这种预测能力可以帮助学生更好更快地学习，这对于构建下一代智能和个性化教育至关重要。

知识追踪是基于学生行为序列进行建模，预测学生对知识的掌握程度。知识追踪是构建自适应教育系统的核心和关键。在自适应的教育系统中，无论是做精准推送，学生学习路径规划或知识图谱构建，第一步都是能够精准预测学生对知识的掌握程度。

——知识追踪-Knowledge Tracing

Part2时间安排

• November 17, 2022 - Start Date.

• December 31, 2022 - Final submission deadline.

• January 2, 2023 - Final competition results announced.

All deadlines are at 11:59 PM UTC on the corresponding day unless otherwise noted. The competition organizers reserve the right to update the contest timeline if they deem it necessary.

Part3奖励机制

• We will provide cash prizes for the top-3 teams (1st place: 600 ; 3rd place $300)

• An official certificate will be awarded to the top-3 teams. The top-3 teams will be invited to give oral presentations during AAAI 2023.

• The first author of the top-3 teams will be invited to contribute to a competition review paper.

Note: The top-3 teams should make their training and testing code publicly available for verification after the testing submission deadline.

Part4赛题描述

The KT related research has been studied since 1990s where Corbett and Anderson, to the best of our knowledge, were the first to estimate students’ current knowledge with regard to each individual knowledge component (KC). A KC is a description of a mental structure or process that a learner uses, alone or in combination with other KCs, to accomplish steps in a task or a problem. Since then, many attempts have been made to solve the KT problem, such as probabilistic graphical models and factor analysis based models. Recently, due to the rapid advances of deep neural networks, deep learning based knowledge tracing (DLKT) models have become the de facto KT framework for modeling students’ mastery of KCs.

However, even through a large body of deep learning based KT models are proposed, the majority of existing baselines don’t utilize the rich auxiliary side information in educational contexts. Various auxiliary side information could be extracted as external knowledge and integrated with the DLKT models. These auxiliary knowledge are expected to improve DLKT performance, which can be considered as follows:

• Question side information: (1) question text content; (2) latent question variations with respect to each KC; (3) question difficulty level; and (4) relations among questions.

• Student side information: (1) historical successful and failed attempts; (2) recent attempts; (3) students’ learning ability; and (4) individualized prior knowledge of students.

• KC side information: (1) latent knowledge representation; and (2) relations among KCs.

Therefore, in this competition, we would like to call for researchers and practitioners to improve the KT models’ performance by considering rich side information. The proposed education challenge will release a large student assessment dataset with rich textual and structural information.

---