Prediction and Evaluation of Students’ Performance in E-Learning Using Data Mining Algorithm

Purpose: The proposed work aims to analyze the used and most efficient machine learning techniques in both online and offline education contexts, for different objectives, this research used various machine learning approaches to predict student performance. The educational data is analyzed based on performance among all compared algorithms on the dataset.

Methods: We used the algorithms k-NN, SVM, LR and ANN for student performance analysis depending on various input features, students from secondary schools and an intermediate school in Iraq were used to compile the performance dataset for the students, the dataset includes attributes suggested about the students, our data set includes 800 instances, with 8 attributes. We used a variety of classification algorithms on datasets of student performance to analyses the dataset and improve the generic reliability of the algorithms to find classifiers with higher accuracy. So, we calculate the algorithm's accuracy, that the accuracy of all implemented classification methods is training and testing calculated by using split-validation. The student's performance dataset was used to train and after that test a variety of classification methods.

Results: ANN classification algorithm's results are very encouraging when compared to other methods for classification. Neural Networks Classification Algorithm, when we experimented with the whole 8 attributes, this algorithm showed 100% as the highest accuracy, and when implementation The logistic regression Classification Algorithm, when we experimented with the whole 8 attributes, this algorithm showed 99.71% as the highest accuracy and when implementing the K Nearest Neighbor KNN Classification Algorithm, when experimented with the whole 8 attributes, this algorithm showed 99.43%, but when we implementation SVM Classification Algorithm, when we experimented with the whole 8 attributes, this algorithm showed 99.17% as the Lower Accuracy compares with other algorithms. Conclusions: Using E-learning classification algorithms for a student performance dataset using data mining methods, data mining algorithms have been applied to the student performance data set in E-learning. ANN classification algorithm's results are very encouraging when compared to other methods for classification, when we apply the training data ANN classifier the result of accuracy is higher, While in SVM little lasting accuracy. We can conclude that ANN is a more efficient model than others for classifying student performance across these multiple attributes. The best important factors that affect student effort have been made to improve the quality of E-learning.