possibility01 / flood-prediction-in-malawi--zindi-competition

Simple starter code

Flood-Prediction-in-Malawi--Zindi-Competition

Starter Code for Flood Prediction in Malawi

Author: Darius Moruri

---

• This is a simple starter code to get you going for the Zindi flood prediction competition
• As this is just a basic machine learning pipeline, the following aspects haven't been covered:
  • Exploratory Data Analysis
  • Feature Engineering
  • Feature Selection
  • Hyperparameter Tuning
  • Model Evaluation
  • Model interpretation
  • Sourcing for more data
  • Documentation and Presentation

Despite its basic approach, this starter code yieldied a satisfacatory RMSE of 0.11866 and a top 15 ranking (as at the time of writing) in the public leaderboard

Context

On 14 March 2019, tropical Cyclone Idai made landfall at the port of Beira, Mozambique, before moving across the region. Millions of people in Malawi, Mozambique and Zimbabwe have been affected by what is the worst natural disaster to hit southern Africa in at least two decades.

In recent decades, countries across Africa have experienced an increase in the frequency and severity of floods. Malawi has been hit with major floods in 2015 and again in 2019. In fact, between 1946 and 2013, floods accounted for 48% of major disasters in Malawi. The Lower Shire Valley in southern Malawi, bordering Mozambique, composed of Chikwawa and Nsanje Districts is the area most prone to flooding.

The objective of this challenge is to build a machine learning model that helps predict the location and extent of floods in southern Malawi.

Data

The training data for this competion can be found here and a sample of the submission file can be found here

Evaluation

The error metric for this competition is the Root Mean Squared Error