Mining And Recognition Of English Learning Patterns For Mobile Users Based On Time Series Clustering And Ensemble Learning

The related research in the field of mobile assisted language learning at home and abroad has gone through nearly 20 years of development and basically follows the route of gradual progress from theory to application practice.However,there are few procedural studies on learners’individual language skills learning behavior based on mobile platform data.Taking vocabulary learning as the breakthrough point,this paper uses time series clustering and integrated modeling method to mine and analyze the user learning behavior data of an English vocabulary learning APP in China.Firstly,taking’single-day memorized word volume’as the measure index,the word-reciting records of users in the whole use cycle are extracted and processed into trajectory data,and the KmL algorithm is used to cluster these trajectories.According to the average trajectory in the class,the characteristics of learning behavior changes of different user groups are summarized,and two learning patterns are depicted.Then,based on the clustering experimental results,the learning pattern prediction model is constructed with the learning state representation attribute as the explanatory variable.The SMOTE-Tomek mixed sampling technique is used to balance the number of different types of samples in the training set;RF-RFE algorithm is used for feature selection to locate the key features in the user learning pattern classification problem;Five single models and five homogeneous ensemble models are constructed,and all base learners are combined by Stacking combination strategy to build a Stacking heterogeneous ensemble model with SVM as meta-classifier.The model evaluation results verify the superiority of the SVM-based Stacking integrated model in user behavior pattern recognition;It also shows that it is reasonable and effective to predict the user’s long-term learning pattern by using the daily learning state representation attributes.Therefore,paying attention to the user learning status reflected behind the behavior data recorded by the mobile English learning platform can help platform developers and operators predict their learning patterns throughout the use cycle,and then design corresponding incentive strategies to improve users ’learning enthusiasm and learning completion.