Build an end-to-end ML system that predicts talent/skill demand across South African industries, and investigates how degree requirements vs. skill-based hiring contributes to the unemployment crisis.
# SA Talent Demand Predictor - ML Engineering Project
# 🎯 Project Goal
Build an end-to-end ML system that **predicts talent/skill demand across South African industries**, and investigates how **degree requirements vs. skill-based hiring** contributes to the unemployment crisis.
> **Core Thesis:** South Africa's unemployment rate is partly driven by degree gatekeeping in job postings — not a shortage of skilled people. This project quantifies that gap and builds a fairer talent-matching system.
>
---
# 📌 Problem Framing
Two interconnected problems to solve:
1. **Predictive Problem** - Which skills will be in demand in which industries over the next 6–24 months?
2. **Analytical / Causal Problem** - Do degree requirements in job postings correlate with longer vacancy periods, hiring discrimination, or unemployment rates in specific demographics?
---
# 🗄️ Data Sources
## South African Government & Official Data
| Source | What You Get | Link |
| --- | --- | --- |
| **Stats SA (QLFS)** | Quarterly Labour Force Survey - unemployment by province, sector, education level | stats.gov.za |
| **DHET** | Graduate output by field, TVET college data | dhet.gov.za |
| **SETA Reports** | Sector Skills Plans with demand forecasts (MERSETA, FASSET, INSETA, etc.) | Each SETA has its own portal |
| **LMIP** | Labour Market Intelligence - occupation & sector data | lmip.org.za |
| **World Bank Open Data** | SA employment, education, and economic indicators | data.worldbank.org |
## Job Postings (Scraping / APIs)
| Source | Method |
| --- | --- |
| **LinkedIn** | API (with approval) or scrape - extract skills vs. degree requirements |
| **Pnet** | Scrape - skills, degree requirements, location, salary |
| **CareerJunction** | Scrape — industry, seniority, job descriptions |
| **Indeed ZA** | Scrape - salary bands, time posted |
## Skill Taxonomy Reference
- **ONET Online** (onetonline.org) - rich skill-to-occupation mappings, adaptable to SA context
- **Kaggle** - search …