philippzi98/food_insecurity_predictions_nlp

Extracts semantic causes of food insecurity from 11.2 million news articles using frame-semantic parsing and causal relation extraction, then validates predictive signals through Granger causality testing before feeding them into a Random Forest regression model. The pipeline combines NLP techniques (seed selection, keyword expansion via semantic similarity) with traditional econometric validation and integrates external food insecurity classifications from FEWS NET. Code is modularized across six reproducible steps, enabling adaptation to custom datasets and use cases beyond the original Factiva news corpus.

No commits in the last 6 months.