Publications by statcompute
Assess Variable Importance In GRNN
Technically speaking, there is no need to evaluate the variable importance and to perform the variable selection in the training of a GRNN. It’s also been a consensus that the neural network is a black-box model and it is not an easy task to assess the variable importance in a neural network. However, from the practical prospect, it is helpful ...
2411 sym 4 img
Merge MLP And CNN in Keras
In the post (https://statcompute.wordpress.com/2017/01/08/an-example-of-merge-layer-in-keras), it was shown how to build a merge-layer DNN by using the Keras Sequential model. In the example below, I tried to scratch a merge-layer DNN with the Keras functional API in both R and Python. In particular, the merge-layer DNN is the average of a multil...
1016 sym 2 img
Partial Dependence Plot (PDP) of GRNN
The function grnn.margin() (https://github.com/statcompute/yager/blob/master/code/grnn.margin.R) was my first attempt to explore the relationship between each predictor and the response in a General Regression Neural Network, which usually is considered the Black-Box model. The idea is described below: First trained a GRNN with the original tra...
2954 sym 2 img
Permutation Feature Importance (PFI) of GRNN
In the post https://statcompute.wordpress.com/2019/10/13/assess-variable-importance-in-grnn, it was shown how to assess the variable importance of a GRNN by the decrease in GoF statistics, e.g. AUC, after averaging or dropping the variable of interest. The permutation feature importance evaluates the variable importance in a similar manner by per...
1443 sym 2 img
GRNN vs. GAM
In practice, GRNN is very similar to GAM (Generalized Additive Models) in the sense that they both shared the flexibility of approximating non-linear functions. In the example below, both GRNN and GAM were applied to the Kyphosis data that has been widely experimented in examples of GAM and revealed very similar patterns of functional relationshi...
1069 sym 2 img
GRNN with Small Samples
After a bank launches a new product or acquires a new portfolio, the risk modeling team would often be faced with a challenge of how to estimate the corresponding performance, e.g. risk or loss, with a limited number of data points conditional on business drivers or macro-economic indicators. For instance, it is required to project the 9-quarter ...
3420 sym
Improve General Regression Neural Network by Monotonic Binning
A major criticism on the binning algorithm as well as on the WoE transformation is that the use of binned predictors will decrease the model predictive power due to the loss of data granularity after the WoE transformation. While talk is cheap, I would use the example below to show that using the monotonic binning algorithm to pre-process predict...
1643 sym
YAP: Yet Another Probabilistic Neural Network
By the end of 2019, I finally managed to wrap up my third R package YAP (https://github.com/statcompute/yap) that implements the Probabilistic Neural Network (Specht, 1990) for the N-category pattern recognition with N > 3. Similar to GRNN, PNN shares same benefits of instantaneous training, simple structure, and global convergence. Below is a de...
1099 sym R (749 sym/1 pcs)