metadata
language: en
tags:
- job-classification
- salary-prediction
- experience-prediction
- deberta
JobPredictor1
Fine-tuned DeBERTa-v3-small model that predicts:
- Expected years of experience required for a job
- Lower salary bound (USD)
- Upper salary bound (USD)
Input Format
[TITLE]: <job title> [DESC]: <job description>
Outputs
| Output | Type | Description |
|---|---|---|
| expected_experience_years | int | Years of experience required |
| pay_lower | int | Lower salary bound (USD) |
| pay_upper | int | Upper salary bound (USD) |
Normalization
Predictions are z-score normalized. Use norm_stats.json to denormalize:
real_value = pred * norm_stats[col]["std"] + norm_stats[col]["mean"]
Test Set Performance
| Metric | Value |
|---|---|
| Experience MAE | 0.57 years |
| Experience Within 1yr | 83.1% |
| Pay Lower MAE | $15,511 |
| Pay Lower Within $20k | 84.5% |
| Pay Upper MAE | $20,190 |
| Pay Upper Within $20k | 76.0% |
Base Model
microsoft/deberta-v3-small