DBSCAN _Algorithm

Description of the Dataset The dataset used in this Python-based DBSCAN exercise contains 300 observations and six numerical variables describing different characteristics of students' academic behavior and performance. The dataset is designed specifically for unsupervised learning, meaning that no predefined cluster label is provided. The objective is therefore to allow DBSCAN to discover groups of students according to the similarity and density of their characteristics. The six variables are Study_Hours, representing the approximate number of hours a student spends studying; Attendance_Percent, representing the percentage of classes attended; Assignments_Completed, indicating the number of assignments completed; Previous_GPA, representing the student's previous grade-point average; Sleep_Hours, indicating the average number of hours of sleep per day; and Exam_Score, representing the student's final examination score. The observations have been constructed to contain different density regions in the data. Students with relatively high study hours, attendance, previous GPA, and exam scores tend to form one group, while students with moderate or lower values tend to form other groups. This structure makes the dataset appropriate for demonstrating the main principle of DBSCAN: identifying dense groups of observations without specifying the number of clusters in advance. Importantly, the dataset does not contain a variable indicating the true cluster membership. The six variables are used as input features, and the cluster labels will be generated by DBSCAN during the analysis. This allows students to examine how the algorithm discovers clusters and identifies observations that may be considered noise or outliers. Before applying DBSCAN, the variables should be examined and, because they are measured on different scales, standardized. This is particularly important because DBSCAN relies on distance calculations, and variables with larger numerical scales could otherwise have a disproportionate influence on the neighborhood structure. This dataset provides a practical transition from the manual DBSCAN exercises to a realistic Python implementation involving a larger number of observations and multiple variables.