Research on Model of Students' Behavior Pattern Recognition Using Big Data and Feature Extraction

In this paper, a student behavior pattern recognition model based on big data and feature extraction is proposed to improve the accuracy and robustness of poor students identification. The study ensures data quality through efficient data acquisition and preprocessing methods, and the dataset contains 1140288 consumption records of 5864 freshmen in the class of 2023. In terms of feature extraction, three dimensions of “average daily consumption amount”, “supermarket consumption amount” and “high consumption proportion” are used, and the PCA dimensionality reduction technique is used to reduce the feature space to two dimensions, so as to facilitate the visualization of clustering results. The improved K-means+ algorithm initializes the cluster center through part of the existing class label information, and introduces the weight parameter wj<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">β</sup> to improve the accuracy and robustness of clustering. The experimental results show that the recognition accuracy of the improved K-means+ algorithm reaches 44.5% in 249 samples, which is significantly better than 40.8% of the traditional K-means algorithm. These results indicate that the proposed model can identify and classify student behaviors more effectively in practice.