mabera/nigeria-poverty-prediction-model

Nigeria Poverty Prediction Model

AutoScientist Challenge 2026 | Finance Category

Author: Hussein Adeiza (mabera)
Role: Licensed Environmental Health Officer, Abuja Nigeria
Base Model: Mixtral 8x7B
Fine-tuned with: AutoScientist by Adaption Labs

Model Description

This is a LoRA adapter fine-tuned on Nigeria state-level Multidimensional Poverty Index (MPI) data. It predicts and explains poverty headcount ratios across Nigeria's 37 states using MPI components and geographic indicators.

Training Data

• Source: OPHI Global MPI 2023 / World Bank / UNDP HDR
• Dataset: 289 adapted rows (Hausa, Yoruba, French, English)
• Quality improvement: 680% (Grade E → B)
• Kaggle: https://www.kaggle.com/datasets/yunusahusseinadeiza/nigeria-poverty-prediction-model-state-level

Training Metrics

• Win rate: 58% adapted vs 42% base model
• Base model: mistralai/Mixtral-8x7B-Instruct-v0.1
• Method: LoRA
• Dataset quality: 1.0 → 7.8 (+680% improvement)

Key Findings

• Northern states average 49.0% poverty headcount vs 10.3% in the South
• North-South gap: 38.7 percentage points
• Most impoverished: Bauchi (75.4%), Jigawa (74.8%), Kebbi (73.8%)
• Least impoverished: Lagos (1.1%), Anambra (2.0%)
• National poverty dropped from 42.3% (2013) → 33.0% (2021)

Why This Matters

Nigeria has over 133 million people living in multidimensional poverty. This model makes state-level poverty prediction accessible to policymakers, development practitioners and financial inclusion specialists working to close Nigeria's inequality gap.

Credits

Powered by Adaptive Data — Adaption Labs
AutoScientist Challenge 2026 | Finance Category