Educational Data Mining: Results from In Vivo Experiments to Teach - PowerPoint PPT Presentation

Educational Data Mining: Results from In Vivo Experiments to Teach Different Physics Topics Advance Technology and Applied Science Research Center National Polytechnic Institute - Mexico Alejandro Ballesteros Román Daniel Sánchez Guzmán AAPT SM 2014, Minneapolis, MN 1

Introduction  Educational Data Mining (EDM) uses different algorithms for analyzing response and behavior in the teaching-learning process for obtaining useful patterns.  These algorithms let to analyze and classify students' behavior or state of knowledge from different concepts. 2

Introduction  Most of these algorithms have not been tested in Physics Education Research.  This work presents the results obtained from applying algorithms used by EDM for teaching different physics concepts applied to in-vivo experiments. EDM Data sets* non-Physics Physics Total 67 27 94 71% 29% 100% * Pittsburgh Science of Learning Center (PSLC - CMU). Data Shop Public Data Sets. 3 https://pslcdatashop.web.cmu.edu/index.jsp?datasets=public

How Educational Data Mining works? 4

Implementation • EDM algorithms (Tree Decision Making, C4.5); were applied with N = 395 students. • Level: High-school students. • Topic: Electric Circuits and Ohm’s Law. • Multiple choice questions: 6 (Academy design, aprox. Electric Circuits Concept Evaluation - ECCE). 5

Results Q3 Q6 Q5 Q4 Q2 Q1 * Did not answer. 6

Results Q3 What does this mean? Q6 Q5 Q4 Q2 Q1 * Did not answer. 7

Results Q3 Most wrong Q6 answered question. Q5 Q4 Q2 Q1 * Did not answer. 8

Results Q3 Q6 Q5 Q4 Q2 Possibly: Confused or not attended concept Q1 during instruction. (C – Correct answer) * Did not answer. 9

Results Q3 Most correctly answered question. Q6 Q5 Q4 Q2 Q1 * Did not answer. 10

Results Q3 Q6 Q5 Q4 Q2 Possibly: Answer is inside the question or well- Q1 attended concept. (C – Correct answer) * Did not answer. 11

Results Q3 Most balanced Q6 question. Q5 Q4 Q2 Q1 * Did not answer. 12

Results Q3 Q6 Q5 Q4 Q2 Possibly: Well-defined question and well- Q1 attended concept. (B – Correct answer) * Did not answer. 13

Results  It is necessary to apply these algorithms with more students for having a fine-grained results.  The use of valid inventories/test would eliminate mistakes.  Patters obtained let us to identify mistakes and wrong questions applied to students, also instruction could be highly improved. 14

Questions? dsanchez@ipn.mx dsanchezgzm@gmail.com 15