| --- |
| license: cc0-1.0 |
| size_categories: |
| - 10M<n<100M |
| --- |
| # Student Performance Dataset |
|
|
| ## Dataset Description |
|
|
| This dataset contains **ten million** synthetically generated student performance records, designed to mimic real-world educational data at the high-school level. It includes detailed demographic, socioeconomic, academic, behavioral, and school-context features for each student, suitable for benchmarking, machine learning, educational research, and exploratory data analysis. |
|
|
| ## File Information |
| | Split | File Name | Number of Records | Approx. Size | |
| |-------------|--------------------------------------|---------------------------|--------------| |
| | **Train** | `train.csv` | 8,000,774 | ~1.23 GB | |
| | **Validation** | `validation.csv` | 999,229 | ~158 MB | |
| | **Test** | `test.csv` | 999,997 | ~158 MB | |
|
|
|
|
| ## Columns & Descriptions |
|
|
| | Column Name | Description | |
| | --------------------- | -------------------------------------------------------------------------------- | |
| | **Age** | Student’s age in years (14–18). | |
| | **Grade** | Grade level (9–12), derived from age. | |
| | **Gender** | Student gender (`Female`, `Male`). | |
| | **Race** | Race/ethnicity (`White`, `Hispanic`, `Black`, `Asian`, `Two-or-more`, `Other`). | |
| | **SES_Quartile** | Socioeconomic status quartile (1 = lowest, 4 = highest). | |
| | **ParentalEducation** | Highest education of parent/guardian (`<HS`, `HS`, `SomeCollege`, `Bachelors+`). | |
| | **SchoolType** | Type of school attended (`Public`, `Private`). | |
| | **Locale** | School location (`Suburban`, `City`, `Rural`, `Town`). | |
| | **TestScore_Math** | Math achievement score (0–100). | |
| | **TestScore_Reading** | Reading achievement score (0–100). | |
| | **TestScore_Science** | Science achievement score (0–100). | |
| | **GPA** | Cumulative Grade Point Average on a 0.0–4.0 scale. | |
| | **AttendanceRate** | Fraction of school days attended (0.70–1.00). | |
| | **StudyHours** | Average self-reported homework/study hours per day (0–4). | |
| | **InternetAccess** | Home internet access (1 = yes, 0 = no). | |
| | **Extracurricular** | Participation in clubs/sports (1 = yes, 0 = no). | |
| | **PartTimeJob** | Holds a part-time job (1 = yes, 0 = no). | |
| | **ParentSupport** | Regular parental help with homework (1 = yes, 0 = no). | |
| | **Romantic** | Currently in a romantic relationship (1 = yes, 0 = no). | |
| | **FreeTime** | Amount of free time after school on a scale from 1 (low) to 5 (high). | |
| | **GoOut** | Frequency of going out with friends on a scale from 1 (low) to 5 (high). | |
| |
| ## Usage |
| This dataset is ideal for: |
| - **Educational Research:** Model how demographics and SES impact academic outcomes. |
| - **Machine Learning:** Train and evaluate predictive models (regression, classification, ordinal) on GPA, test scores, or attendance. |
| - **Clustering & Segmentation:** Identify student subgroups (e.g., high-achievers, at-risk) for targeted interventions. |
| - **Fairness Analysis:** Examine performance disparities across sensitive groups (race, gender, SES). |
| - **Policy Simulation:** Estimate the effects of interventions (e.g., increased study time, universal internet access). |
| |
| ## Example Workflows |
| 1. **Supervised Regression:** Predict GPA from study hours, attendance rate, and parental education. |
| 2. **Classification:** Identify students at risk of chronic absenteeism (AttendanceRate < 0.90). |
| 3. **Clustering:** Segment students into performance-based clusters for personalized learning paths. |
| 4. **Bias Mitigation:** Compare model performance across race/ethnicity groups and apply fairness techniques. |
| |
| ## Data Preprocessing Tips |
| - **One-Hot Encoding:** For categorical features (`Gender`, `Race`, `SchoolType`, `Locale`, `ParentalEducation`). |
| - **Ordinal Encoding:** Map `SES_Quartile`, `FreeTime`, `GoOut` directly to integers. |
| - **Scaling:** Standardize continuous features (`TestScore_*`, `GPA`, `AttendanceRate`, `StudyHours`). |
| - **Shuffle Splits:** Although splits are random, consider re-shuffling training data before each epoch. |
| |
| ## License |
| This dataset is entirely synthetic and contains no real personal data. It is released under the [CC0 1.0 Universal](https://creativecommons.org/publicdomain/zero/1.0/) license for any research, educational, or commercial use. |
| |
| ## Acknowledgements |
| Generated to support reproducible educational data science and machine-learning research. Please cite this project if you use the data in your work. |
