From Learning Styles to Learning Behaviors

Abstract

This repository contains the Jupyter Notebook developed to perform the complete data analysis and machine learning workflow for behavioral clustering in adaptive learning. The notebook implements data preprocessing, exploratory analysis, clustering, statistical validation, and visualization to identify meaningful learner profiles from educational data. The analysis includes the following main components: Data Preparation: Loading the dataset, inspecting data quality, handling missing values and duplicates, and preparing the variables for analysis. Data Preprocessing: Encoding categorical variables. Feature scaling and normalization. Data transformation for machine learning algorithms. Exploratory Data Analysis (EDA): Descriptive statistics. Correlation analysis. Distribution analysis. Visualization of behavioral and academic variables. Clustering Analysis: Applying the K-Means clustering algorithm. Determining the optimal number of clusters using the Elbow Method and Silhouette Score. Evaluating clustering quality using Silhouette Score and Davies-Bouldin Index. Dimensionality Reduction & Visualization: Using Principal Component Analysis (PCA) to visualize learner clusters and explore feature contributions. Statistical Analysis: Performing statistical tests and regression analyses to evaluate differences between the identified learner profiles. Interpretation: Characterizing each cluster based on behavioral patterns and academic performance, providing insights for adaptive learning and educational data mining applications. The notebook was implemented entirely in Python using Jupyter Notebook and relies on widely used scientific libraries, including NumPy, pandas, Matplotlib, SciPy, scikit-learn, and StatsModels. It is shared to promote transparency, reproducibility, and reuse in Learning Analytics, Educational Data Mining, and Artificial Intelligence in Education research.