Blog 7

As the end of summer sessions on campus approaching, we are keep making progress on the research but meanwhile facing more and more difficulties. To processing HBM in R is one of them. Since Hierarchical Bayesian Model is still on its way to develop, it is kind of hard and time-consuming to find some specific functions that fit right into our model. There are so many arguments and different functions in the package to explore. While doing more and talking more to the professor about it, I also start on learning about the validation part of random forest.

The most common technical validations will have two major parts. The first one would be the root mean squared error (RMSE), percent root mean squared error (RMSE percent), and the mean absolute error (MAE). To execute it, we have to aggregate population counts for each selected country within the next coarser administrative level boundary than the finest for which they were available. And then, compare with the observed census counts within each unit. The other one would be out-of-bag (OOB) error estimation, which is also a prominent feature of random forest. It is internally calculated during the RF model fitting and can be considered a robust and unbiased measurement of the prediction accuracy of the model itself. (Some of the information about validation explanation is authentic and online) I tried to use some functions to examine the plot we made and for now it has a relative significant error to adjust. I suppose that might largely because we did not download and include enough variables to run the prediction function.

To research more about HBM and implement it and to improve the accuracy of the prediction from random forest are our two primary goals about this research. And professor Frazier and I will keep working on it during the rest of the summer and the beginning of next semester.

Speak Your Mind