Kathmandu Rain Prediction
Rainfall forecasting in Kathmandu using multi-year weather data, temporal and geographic features, and comparative machine-learning and deep-learning models. Related research was listed at ICETID 2026.
Local weather. A question worth modeling.
- Temporal and geographic feature engineering
- Comparative ML and deep-learning models
- Related ICETID 2026 publication
The problem and its constraints.
This project investigates rainfall forecasting for Kathmandu using multi-year weather data. The supplied CV describes the pipeline and evaluation; the public repository provides the associated project files. Its related publication is co-authored with Saurav Katwal and Rajad Shakya.
- Weather observations contain temporal structure.
- Geographic and temporal features influence model inputs.
- Different model families need comparable evaluation.
What the evidence supports.
Shaswot's CV describes work on the forecasting pipeline, feature engineering, and model comparison. It also lists him as a co-author of the ICETID 2026 publication. The available evidence does not assign individual paper sections or exclusive ownership.
Inside the approach.
Represent the time and place
The CV describes temporal and geographic features derived from multi-year weather data. These features establish the input representation used by the forecasting models.
Compare model families
Bi-LSTM and GRU models are evaluated alongside Random Forest, SVR, and LightGBM. This spans recurrent deep-learning methods and established machine-learning approaches.
Keep the result in context
The CV reports a best RMSE of approximately 3.79. The supplied summary does not specify the units, test split, or exact winning model, so the portfolio does not add those details.
Results with context.
Best reported RMSE
CV-reported rainfall forecasting result; units and test split are not specified in the supplied CV.
The available CV describes model comparison and evaluation but does not document automated tests or a deployment. The case study therefore focuses on the research pipeline and the reported result.
What this project demonstrates.
The project connects a local forecasting question to feature engineering and comparative modeling, with a related collaborative publication.
Documenting the temporal split, units, baselines, and reproducibility steps would make future comparisons more useful. These are suggested documentation improvements.