Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Personalized Forgetting Mechanism with Concept-Driven Knowledge Tracing

Shanshan Wang, Ying Hu, Xun Yang, Zhongzhou Zhang, Keyang Wang, Xingyi Zhang

TL;DR

This work tackles knowledge tracing under personalized forgetting by introducing CPF, a concept-driven framework that couples individualized learning gains with a causal forgetting mechanism grounded in precursor–successor knowledge relations. CPF adds a personalized learning module to compute student-specific abilities, and a causal forgetting module that leverages a P-matrix to model prerequisite relationships and a forgetting-review process to reflect memory reactivation. The approach uses a prediction module to forecast next-task performance and a cross-entropy objective optimized with Adam. Across three public KT datasets, CPF consistently outperforms state-of-the-art baselines, with ablations confirming the value of personalization, hierarchical concept relations, and forgetting dynamics. The work offers improved interpretability and potential for lifelong learning by accurately simulating how students acquire and forget knowledge over time.

Abstract

Knowledge Tracing (KT) aims to trace changes in students' knowledge states throughout their entire learning process by analyzing their historical learning data and predicting their future learning performance. Existing forgetting curve theory based knowledge tracing models only consider the general forgetting caused by time intervals, ignoring the individualization of students and the causal relationship of the forgetting process. To address these problems, we propose a Concept-driven Personalized Forgetting knowledge tracing model (CPF) which integrates hierarchical relationships between knowledge concepts and incorporates students' personalized cognitive abilities. First, we integrate the students' personalized capabilities into both the learning and forgetting processes to explicitly distinguish students' individual learning gains and forgetting rates according to their cognitive abilities. Second, we take into account the hierarchical relationships between knowledge points and design a precursor-successor knowledge concept matrix to simulate the causal relationship in the forgetting process, while also integrating the potential impact of forgetting prior knowledge points on subsequent ones. The proposed personalized forgetting mechanism can not only be applied to the learning of specifc knowledge concepts but also the life-long learning process. Extensive experimental results on three public datasets show that our CPF outperforms current forgetting curve theory based methods in predicting student performance, demonstrating CPF can better simulate changes in students' knowledge status through the personalized forgetting mechanism.

Personalized Forgetting Mechanism with Concept-Driven Knowledge Tracing

TL;DR

This work tackles knowledge tracing under personalized forgetting by introducing CPF, a concept-driven framework that couples individualized learning gains with a causal forgetting mechanism grounded in precursor–successor knowledge relations. CPF adds a personalized learning module to compute student-specific abilities, and a causal forgetting module that leverages a P-matrix to model prerequisite relationships and a forgetting-review process to reflect memory reactivation. The approach uses a prediction module to forecast next-task performance and a cross-entropy objective optimized with Adam. Across three public KT datasets, CPF consistently outperforms state-of-the-art baselines, with ablations confirming the value of personalization, hierarchical concept relations, and forgetting dynamics. The work offers improved interpretability and potential for lifelong learning by accurately simulating how students acquire and forget knowledge over time.

Abstract

Knowledge Tracing (KT) aims to trace changes in students' knowledge states throughout their entire learning process by analyzing their historical learning data and predicting their future learning performance. Existing forgetting curve theory based knowledge tracing models only consider the general forgetting caused by time intervals, ignoring the individualization of students and the causal relationship of the forgetting process. To address these problems, we propose a Concept-driven Personalized Forgetting knowledge tracing model (CPF) which integrates hierarchical relationships between knowledge concepts and incorporates students' personalized cognitive abilities. First, we integrate the students' personalized capabilities into both the learning and forgetting processes to explicitly distinguish students' individual learning gains and forgetting rates according to their cognitive abilities. Second, we take into account the hierarchical relationships between knowledge points and design a precursor-successor knowledge concept matrix to simulate the causal relationship in the forgetting process, while also integrating the potential impact of forgetting prior knowledge points on subsequent ones. The proposed personalized forgetting mechanism can not only be applied to the learning of specifc knowledge concepts but also the life-long learning process. Extensive experimental results on three public datasets show that our CPF outperforms current forgetting curve theory based methods in predicting student performance, demonstrating CPF can better simulate changes in students' knowledge status through the personalized forgetting mechanism.
Paper Structure (29 sections, 25 equations, 7 figures, 3 tables)

This paper contains 29 sections, 25 equations, 7 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Knowledge Tracing
  4. Forgetting Curve
  5. Preliminaries
  6. Relation Definition
  7. Problem Statement
  8. Methodology
  9. Overview of LPKT
  10. Proposed Method
  11. Personalized Learning Module
  12. Causal Forgetting Module
  13. Predicting Module
  14. Objective Function
  15. EXPERIMENTS
  16. ...and 14 more sections

Figures (7)

  • Figure 1: Examples of three learners answer a series of exercises on an online learning system. (a) indicates that the forgetting rate of various students is different. (b) represents different forgetting patterns: the traditional model considers correct answers as mastering the knowledge point, while incorrect answers indicate no impression of the knowledge point; the time decay model believes that memory of knowledge points weakens over time; and the causal forgetting model integrates the intrinsic relationship between time and knowledge points, balancing long-term and short-term memory.
  • Figure 2: The main network structure of CPF model is mainly composed of learning, forgetting and prediction modules. The learning module uses the learning gate to get the personalized learning gain, uses the directed relationship of knowledge concepts to capture the causality of forgetting process, and finally predicts the future performance of students.
  • Figure 3: Illustration of the mastery level of student knowledge points. On the left side are the individual differences in students' abilities, while on the right side are the mastery levels of knowledge points after students undergo a consistent answering process. (a) and (b) represent the initial ability distribution of students and the degree of mastery of knowledge concepts after the same answering sequence respectively.
  • Figure 4: The evolution of students' knowledge state. The left side represents the students' answer order and results, and the right side describes the updating of the knowledge state due to the knowledge concept relationship.
  • Figure 5: Comparison of exercise-concept correlation study
  • ...and 2 more figures