Neural Network Summary
With best experiments, provides correct placement for about 90% of incoming students
(various approaches yield 80%-90% accuracy)
Further training possible on new data sets, if faculty wish to adjust outcomes
System supports letters to incoming students and Web-based system
System is scaleable and applicable to range of subjects
(as long as a training set is available)
Even if error rate small (e.g., perhaps 10%), no way to know whom those students are.
System is not "fail safe"
created 28 September 2017
last revised 28 September 2017