Figure 5
From: Infinite mixture-of-experts model for sparse survival regression with application to breast cancer
Comparison to a global model. Left: The actual number of points in the test set which scored better in a particular model (442 for MOE Vs 58 for Single Cluster) based on the likelihood scores. Results of the Kruskal-Wallis rank test also validated this observation with a p-value ≪ 0.001. Right: Results of the key interactions found for a single cluster model. Some of the key factors are not identified along with existence of many false-positives.