Research And Application Of Online Learning Behavior Based On Data Mining

With the advancement of the Internet and big data technology,online learning platforms are developing rapidly and the scale of the platforms is expanding.Learners have generated a huge amount of behavioral data in the learning process,and it is possible to understand different learners’ learning situations and patterns by mining these behavioral data.Different behavior leads to different learning effects.According to behavioral science and behaviorist psychology theory,systematically and comprehensively collect online learning behavior characteristics and construct an online learning behavior model.The learning behavior model is used to predict the effect of online learning,so as to achieve the purpose of self-monitoring by learners,timely intervention by teachers,and ultimately improving the learning effect.The main work includes the following aspects: First,supported by behavioral science and behaviorist psychology theories,comprehensive and perfect online learning behavior features are selected and extracted,defined,described,and feature values selected.Online learning behavior features are classified according to the six elements of online learning behavior.The attribute-oriented induction method quantifies the collected behavioral features,gives the basis for threshold selection,and provides solutions for the special cases that may occur for some of the features.Secondly,the correlation analysis between the collected behavioral features and the final learning effect is conducted,including single-factor analysis and multi-feature analysis.To obtain the correlation between the features and the learning effect,present the results visually,select the features with a stronger correlation with the final learning effect,and lay the foundation for building the online learning behavior model.Finally,the initial weights are randomly generated using the BP neural network to output the final prediction effect;the weights of the features are obtained using principal component analysis,and the weights are used as the initial weight input of the BP neural network to predict the online learning effect.The results obtained from the two methods are evaluated using commonly used evaluation metrics to compare the prediction effects.The study shows that: 1.Most behavioral characteristics are positively correlated with final learning outcomes,with some of them being strongly correlated.The minimum superior level of performance on each trait met 60% for better-rated learners and 80% for worse-rated learners on each trait.2.The initial weights are randomly generated using BP neural network to predict the online learning effect;the feature weights are obtained using principal component analysis and put into BP neural network to predict the online learning effect,both of which have an accuracy rate of over 90%,and the comparison of the experimental results of the two can be obtained that the latter has a better prediction effect.The prediction results of the two methods are better,indicating that the modeling process for online learning behavior is more reasonable and that constructed online learning behavior model is effective.