I adapted the original Bayesian Knowledge Tracing (Brute Force) BKT-BF code (www.columbia.edu/~rsb2162/BKT-BruteForce.zip) to allow data in chronological order but not necessary sorted on Skill and Student (can be mixed). Moreover, the dataset can have free number of columns in any order and any separator (coma, tab, etc.)
I didn't tested it thoroughly but at least it works for my needs.
Parameters (1 is the first one):
skillPos = Skill column position
studentPos = Student column position
rightPos = Right column position
separator = Separator (comma(,) , tab(tab), etc.)
$> java ComputeKTparamsAll ./data.txt 4 3 6 tab
In this example, it gets the model from the data.txt dataset where the skill column is the 4th, student column is the 3rd, right column is the 6th and the columns are separated by tab.
Muy buena información, gracias!
ReplyDelete