Learning State Analysis Of Students Based On Outlier Detection

In our country, Educational data mining is still in the phase of theoretical exploration, most of the research is theoretical description and feasibility analysis, applied research very little. This paper explores the application of data mining technology in the learning state analysis of students based on the research topic, from the perspective of applied research of the educational data mining. In the presence of college students and student supervisors imbalance problem in Chinese universities, leading to lack of effective use of educational management resources. This paper is intended to solve a practical problem: Assign to which students the scarce educational management resources of personalized learning guidance, to help them successfully graduate, so that the limited educational management resources to play a greater value. This paper use learning state analysis of students based on outlier detection to find students with exceptional learning state, to provide the basis for the assignment of educational management resources.To recognize students who have exceptional examination grades data, a hybrid two-stage outlier detection algorithm is designed for this problem. This algorithm uses a density-based outlier detection algorithm to calculate every student's local outlier factor, then uses a statistics-based outlier detection algorithm to make either-or judgments, so that to recognize students who have exceptional examination grades data. Vast majority of them have exceptional learning state, but the total number is below normal. Some improvements are achieved for the lack of the original algorithm, from the perspective of the data, artificial attributes are added to expand the knowledge base; from the perspective of the algorithm, reduced iteration is applied to iteratively detect the decreasing data set; finally, the two improvements fuse to form a comprehensive improvement. In this problem, the loss of a student with exceptional learning state costs too much, Compared to the original algorithm, the improved algorithms find more students of exceptional learning state and realize the optimization on the quantity of students with exceptional learning state. It turns out that on the condition of using the hybrid two-stage outlier detection algorithm or improved algorithms, learning state analysis of students based on outlier detection can effectively recognize students with exceptional learning state, the results can assist College student supervisors more scientific and efficient management.