| International students planing to study in United States colleges and universities must demonstrate English language proficiency before gaining university admission. Many participate in an intensive English program as a means to gain such proficiency, but some learners demonstrate consistent progress through the program whereas others are less successful.;Selected features were then applied using the Outcome Advisor in a classification mode, with predicted category of success based on the patterns of high, medium, and low success clusters found in the knowledge bank training sample. Results indicated that selected features could be used by the OA to accurately classify learners into success categories, averaging 64% accuracy when judging three success categories (high, medium, low) and 80% accuracy with two success categories (high-medium versus low).;This study examined variability in learner progress in relationship to 49 language learning (LL) background variables. A knowledge base was developed consisting of background and program progress data for individual learners. The Outcome Advisor (R) knowledge base system (OA), which uses statistical pattern recognition and artificial intelligence for complex deductive processes, was used to build the knowledge bank and to select those features which most clearly discriminated between high, medium and low success learners. Feature refinement resulted in a set, or taxonomy, of features which must clearly differentiated between the success categories. In addition, "profiles" for high, medium, and low progress (i.e. success) groups were constructed by identifying feature values which were most clearly associated with each success group. Comparisons were also made to a foundation study examining LL background features associated with entry TOEFL scores. |
