LSTM, Transformer, and LLM NLP regression
LSTM, Transformer and LLM regression models to predict a hotel’s score by review | LSTM Code, Transformer Code, LLL code
- 8.13% error for LSTM
- 7.38% error for fine-tuned Transformer
- 7.29% error for fine-tuned LLM
Bayesian models for A/B testing
| Methodology and analysis for Bayesian Two-Step Aggregated model, Dirichlet-Multinomial Model, Dirichlet-Poisson Mixture Model for estimating absolute lift samples given treatment and control groups conversion histograms. | Python Code |
GNN nodes prediction model
Caste prediction for households using social network graphs and GNN. | Python Code
- 75% accuracy GNN model that predicts belonging to one of five castes for unbalanced data with social network graph structure
DML and Causal Forest (with Paul Delatte)
Re-estimating B. Jones & B. Olken (2009) paper using DML and Causal Forest models. | Python Code
- Estimations using DML PLR and IRM models from DoubleML with RF and finetuned LGBM classifiers + Causal Forest from econml.dml