Datasets:
| license: mit | |
| task_categories: | |
| - tabular-classification | |
| tags: | |
| - tabular | |
| - classification | |
| - student-grades | |
| - label-errors | |
| - noisy-labels | |
| - data-centric-ai | |
| # Student Grades Demo Dataset | |
| This dataset contains student grades data with both true labels and noisy (corrupted) labels. | |
| ## Dataset Description | |
| The dataset includes: | |
| - Student exam scores (exam_1, exam_2, exam_3) | |
| - Notes field | |
| - True letter grades (`letter_grade`) | |
| - Noisy/corrupted letter grades (`noisy_letter_grade`) | |
| This is useful for demonstrating and validating label error detection methods. | |
| ## Usage | |
| ```python | |
| import pandas as pd | |
| # Load the dataset | |
| df = pd.read_csv("hf://datasets/Cleanlab/student-grades-demo/student-grades-demo.csv") | |
| print(df.head()) | |
| # Find rows where labels were corrupted | |
| import numpy as np | |
| true_errors = np.where(df["letter_grade"] != df["noisy_letter_grade"])[0] | |
| print(f"Number of label errors: {len(true_errors)}") | |
| ``` | |
| ## License | |
| MIT License | |