Relaxed Clustered Hawkes Process for Student Procrastination Modeling in MOOCs

Hawkes processes have been shown to be efficient in modeling bursty sequences a variety of applications, such as finance and social network activity analysis. Traditionally, these models parameterize each process independently assume that the history point can fully observed. Such could however inefficient or even prohibited certain real-world field education, where assumptions are violated. Motivated by problem detecting predicting student procrastination students Massive Open Online Courses (MOOCs) with missing partially observed data, this work, we propose novel personalized model (RCHawkes-Gamma) discovers meaningful behavior clusters jointly learning all simultaneously, without relying on auxiliary features. Our experiments both synthetic education datasets show RCHawkes-Gamma effectively recover their temporal dynamics, resulting better predictive performance future activities. further analyses learned parameters association delays discovered unveil representations various behaviors students.