Data Mining

Predicting the local epidemics of dengue fever, the competition hosted by DrivenData, Phase one of the analysis aims to examine different models and their predictions using Weka Software, A golden nugget of Dengue predictions for each two cities, San Juan and Iquitos, for year and weekofyear in the test set was best found from the weka RandomSubSpace Model with RandomForest as base classifier which gives a mean absolute error of 6.1411, still there are chances to tune with more models to get better predictions. Phase two of the work explore neural networks using javanns software, using backpropagation multilayer perceptronCS in Weka gave a model error of 8.0933 whereas javanns neural networks gave slighlty better MAE of 7.9332, in the analysis section we will see the examination of the results. Current leaderboard rank in the challenge is at 583 out of 4720 participants.