Boicu, M., & Lalwani, J., & Daga, A. (2023, March), Analyzing Student Procrastination to Identify At-Risk Behavior  Paper presented at ASEE Southeast Section Conference, Arlington, Virginia. 10.18260/1-2--44985

\bibitem{asee_peer_44985}
  Boicu, M., \& Lalwani, J., \& Daga, A. (2023, March), \emph{Analyzing Student Procrastination to Identify At-Risk Behavior}
  Paper presented at ASEE Southeast Section Conference, Arlington, Virginia.
  10.18260/1-2--44985

Mihai Boicu, Jay Lalwani, and Aditya Daga.  "Analyzing Student Procrastination to Identify At-Risk Behavior".  ASEE Southeast Section Conference, Arlington, Virginia, 2023, March.  ASEE Conferences, 2023.  https://peer.asee.org/44985 Internet.  16 Dec, 2024

\bibitem{asee_peer_44985}
  Mihai Boicu, Jay Lalwani, and Aditya Daga.
  "Analyzing Student Procrastination to Identify At-Risk Behavior".
  \emph{ASEE Southeast Section Conference, Arlington, Virginia, 2023, March}.
  ASEE Conferences, 2023.
  https://peer.asee.org/44985 Internet.  16 Dec, 2024

@INPROCEEDINGS{asee_peer_44985
author = "Mihai Boicu, Jay Lalwani, and Aditya Daga"
title = "Analyzing Student Procrastination to Identify At-Risk Behavior"
booktitle = "ASEE Southeast Section Conference"
year = "2023"
month = "March"
address = "Arlington, Virginia"
publisher = "ASEE Conferences"
note = {https://peer.asee.org/44985}
number = {10.18260/1-2--44985}
}

TY  - CPAPER
AB  - Researchers have long considered procrastination as a form of self-destructive behavior and a key factor in university students’ failure. However, procrastination by itself may have positive impact to some students, so, more factors must be taken in account to identify at-risk behavior. In order to predict student failure and allow early warning, we developed and trained a neural network based on second week activities that allow multiple attempts in a college course at George Mason University. The following procrastination related inputs were computed: average time spent between a student’s first and last attempt through all assignments, average time remaining before each assignment deadline from the last attempt, average assignment grade, and average number of assignment attempts. The neural network consists of an input layer of 4 perceptrons, 6 hidden layers, and a single output perceptron. We trained it with each student’s corresponding final grade (pass or fail determined by a grade above or below 59%). Our neural network had an accuracy of 91.6%, precision of 89.9%, and recall of 99.5%. Those who were misclassified largely consisted of passing students with a grade near the passing threshold. Additionally, using permutation feature importance, we found that the average time remaining before assignment deadline was the most important feature in the neural network resulting in an 18% decrease in classification accuracy when its value was randomized, as students who tended to submit assignments late consistently failed the course. Because procrastination is so common in university students, our model is relevant and an appreciable tool for professors to use to try to minimize failure in their class. By warning students of their failure behavior earlier in the course, they will be compelled to address their procrastinative tendencies and greatly reduce their chances of failing.
AU  - Mihai Boicu
AU  - Jay Lalwani
AU  - Aditya Daga
CY  - Arlington, Virginia
DA  - 2023/03/12
PB  - ASEE Conferences
TI  - Analyzing Student Procrastination to Identify At-Risk Behavior
UR  - https://peer.asee.org/44985
DO  - 10.18260/1-2--44985
ER  -