Predicting of project spending for an NGO

Field projects for an anonymous NGO. For this purpose Machine Learning models are used. Data as to what kind of project is planned and the planned spending is used as input for these models. Key problems are overfing,heteroskedasticity, sparse data and a lack of predictive features. We find it not yet possible to be more accurate than the budgets are, although we do think this is possible in the future. To this end we provide some advice to the NGO. The best performing model was the K Nearest Neighbour. Experiments show that with additional computational power Neural Networks can outperform both the K Nearest Neighbour models as well as the budget.