TY - GEN
T1 - Instructional factors analysis
T2 - 4th International Conference on Educational Data Mining, EDM 2011
AU - Chi, Min
AU - Koedinger, Kenneth
AU - Gordon, Geoff
AU - Jordan, Pamela
AU - VanLehn, Kurt
PY - 2011
Y1 - 2011
N2 - In this paper, we proposed a new cognitive modeling approach: Instructional Factors Analysis Model (IFM). It belongs to a class of Knowledge-Component- based cognitive models. More specifically, IFM is targeted for modeling student's performance when multiple types of instructional interventions are involved and some of them may not generate a direct observation of students' performance. We compared IFM to two other pre-existing cognitive models: Additive Factor Models (AFMs) and Performance Factor Models (PFMs). The three methods differ mainly on how a student's previous experience on a Knowledge Component is counted into multiple categories. Among the three models, instructional interventions without immediate direct observations can be easily incorporate into the AFM and IFM models. Therefore, they are further compared on two important tasks-unseen student prediction and unseen step prediction-and to determine whether the extra flexibility afforded by addi- tional parameters leads to better models, or just to over fitting. Our results suggested that, for datasets involving multiple types of learning interventions, dividing student learning opportunities into multiple categories is beneficial in that IFM out-performed both AFM and PFM models on various tasks. However, the relative performance of the IFM models depends on the specific prediction task; so, experimenters facing a novel task should engage in some measure of model selection.
AB - In this paper, we proposed a new cognitive modeling approach: Instructional Factors Analysis Model (IFM). It belongs to a class of Knowledge-Component- based cognitive models. More specifically, IFM is targeted for modeling student's performance when multiple types of instructional interventions are involved and some of them may not generate a direct observation of students' performance. We compared IFM to two other pre-existing cognitive models: Additive Factor Models (AFMs) and Performance Factor Models (PFMs). The three methods differ mainly on how a student's previous experience on a Knowledge Component is counted into multiple categories. Among the three models, instructional interventions without immediate direct observations can be easily incorporate into the AFM and IFM models. Therefore, they are further compared on two important tasks-unseen student prediction and unseen step prediction-and to determine whether the extra flexibility afforded by addi- tional parameters leads to better models, or just to over fitting. Our results suggested that, for datasets involving multiple types of learning interventions, dividing student learning opportunities into multiple categories is beneficial in that IFM out-performed both AFM and PFM models on various tasks. However, the relative performance of the IFM models depends on the specific prediction task; so, experimenters facing a novel task should engage in some measure of model selection.
UR - http://www.scopus.com/inward/record.url?scp=84857471334&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84857471334&partnerID=8YFLogxK
M3 - Conference contribution
AN - SCOPUS:84857471334
SN - 9789038625379
T3 - EDM 2011 - Proceedings of the 4th International Conference on Educational Data Mining
SP - 61
EP - 70
BT - EDM 2011 - Proceedings of the 4th International Conference on Educational Data Mining
Y2 - 6 July 2011 through 8 July 2011
ER -