PREFECT - prediction models for famine catastrophes

Almost all historic as well as current famine catastrophes worldwide occurred due to environmental reasons such as drought, flood, erosion, and other weather phenomena like El Nino in combination with human-made influences such as military and religious conflicts, social economic exploitation and corruption.

The international and private aid organizations are doing their best trying to solve the problems famine-induced mass migration brings along.

The biggest problem aid organizations are facing is the lack of time for consistent and sustainable planning (setting of refugees camps including food and energy supply, infrastructure, security, etc.) prior to its occurrence. Hence, increasing the lead time for preparation is an essential step and will lead to save hundred thousands of lives.

The aim of the project is to increase the lead time by developing a mathematical prediction model that is able to compute the probability of a famine catastrophe by learning from data. In order to perform such computations, the prediction model is developed and trained on historic data of drought, flood, erosion as well as socio-economic data sources, taken from African countries mainly. The project includes several steps: First, data sources need to be identified, prepared (filtered and cleaned) and evaluated. In a second step, a mathematical prediction model has to be set up using common data analytics methods. Then the model has got to be trained by the data collected in the first step. In a third step the prediction accuracy of the model has got be cross-validated. As soon as the model’s accuracy enables a good prediction on the training data sets, real predictions of the probability for famine catastrophes will be computed.