Paper — APA Citation

Predictive Modeling of Typhoid Incidence in Nepal · Master’s Thesis

Thesis · Master’s · Paper landing

Contents

§	Section
1	Introduction — problem, research question, conceptual framework
2	Literature Review — typhoid epidemiology, climate drivers, ML in climate-health
3	Methodology — study design, features, train/test, four models, evaluation
4	Results — predictor structure, model performance, figures, geography
5	Discussion — climate–typhoid nexus, early warning, NAP alignment, six policy recommendations
6	Data — HMIS · CHIRPS · ERA5-Land · DRR portal · CBS shapefiles
7	Sources — full APA reference list & how to cite

Abstract

Typhoid fever imposes a persistent and disproportionate public-health burden on Nepal. This thesis develops, trains, and rigorously evaluates four machine-learning models — Random Forest, XGBoost, Multilayer Perceptron, and a Weighted Ensemble — for the prediction of monthly district-level typhoid case counts across Nepal’s 77 districts and 108 months (January 2015 – December 2023; N = 7,327 district-month observations). Four validated data sources are integrated: HMIS outpatient surveillance, DRR-portal flood records, CHIRPS precipitation, and ERA5-Land temperature and humidity. The Weighted Ensemble achieves the strongest predictive performance across all metrics (R² = 0.8675; RMSE = 126.20; MAE = 68.07; MAPE = 41.01 %) on the held-out 2022–2023 test set. Feature-importance analysis identifies the autoregressive cases_lag1 term as the dominant predictor (normalized importance = 0.670), followed by lagged temperature (0.088), cyclical month encoding (combined: 0.122), and the monsoon indicator (0.037). The study provides actionable evidence for climate-informed typhoid early-warning systems aligned with Nepal’s National Adaptation Plan 2021–2050.

Companion manuscript

The condensed companion manuscript — a tightened three-model comparison anchored on XGBoost as the operational headline — is indexed separately under the Research Manuscript publication.