Ordinal Learning Algorithm With Applications In Final Exam Prediction

Ordinal Learning(OL)is a special supervised learning method in machine learning.It can effectively use the equivalence relationship between data and is widely concerned by researchers.Although the Linear Discriminant Learning for Ordinal Regression(LDLOR)can solve the classification or regression problem of ordered data effectively,it fails to utilize the structural information hidden in the data.With regards to this reason,our paper carries out research focusing on the ordinal learning of embedding local structure preservation and adaptive adjustment of intra-class dispersion The follows are our main work.1.Propose an ordinal regression learning algorithm based on local preservation(LSP-LDLOR)and an ordinal regression learning algorithm based on adaptive local preservation(SaLSP-LDLOR).Based on Image Euclidean Distance with Linear Discriminant Learning for Ordinal Regression(IMED-LDLOR),the LSP-LDLOR algorithm embeds the local structure information of the sample space.Further,the locality preserving matrix is adaptively adjusted to obtain an optimized locality preserving matrix.The locally maintained ordinal regression learning algorithm SaLSP-LDLOR.The LSP-LDLOR algorithm and SaLSP-LDLOR algorithm are corrpared with IMED-LDLOR algorithm on JAFFE and FG-NET facial expression data sets.The experimental results show that the newly developed algorithm is effective.2.Present an ordinal regression learning algorithm based on adaptive intra-class dispersion(SaSw-LDLOR)and an ordinal regression learning algorithm based on local-preserving adaptive intra-class dispersion(LSaSw-LDLOR).Based on IMED-LDLOR model,an ordinal regression learning algorithm based on adaptive intra-class dispersion is constructed by adjusting the intra-class dispersion matrix with fuzzy self-adapting.Furthermore,an ordinal regression learning algorithm LSaSw-LDLOR based on locality preserving adaptive intra-class dispersion is designed by embedding local structure preservation informatioa SaSw-LDLOR and LSaSw-LDLOR are compared with IMED-LDLOR on JAFFE and FG-NET facial expression datasets.The experimental results show the effectiveness of the lately devised algorithm.3.Using the newly designed ordinal learning algorithm to predict the final examThe applicability of the newly designed algorithm was verified by the prediction of the university computer basic course results of a college of Nanjing Normal University in 2017.The training set comes from the students’ classroom attendance,card consumption,library access,internet time and duration,and process assessment.The accuracy of the classic machine learning algorithm and the ordered regression learning algorithm for the final grade prediction of the sample students are analyzed.The practicality of the newly proposed algorithm is verified in the performance prediction experiments.