Evaluating my models

oneponep Member Posts: 20 Maven
edited November 2018 in Help



Im trying to evalute my models - I've build 3 models, a k-NN, a local polynomial regression and a deep learning.

I'm trying do predict the shaftpower of a vessel with different variations of trim. I've put it all in this chart.

Plot 4 (7).png

I've made a confidence line made from the k-NN anomaly score - so it measures the distance from the point and to the nearest points to tell how confident the model is in its prediction. However, it seems like my models are not really confident in predicting anything away from -0,1, because the confidence level begins the fall considerably.


How could I improve this? Is it because there is not enough data?




  • Options
    Thomas_OttThomas_Ott RapidMiner Certified Analyst, RapidMiner Certified Expert, Member Posts: 1,761 Unicorn

    Well it's hard to say what's going on without seeing your data and a process. How much data are you processing and have you tried other algorithms?  There's probably some optimziation you can do for the "k" value to make this better.

Sign In or Register to comment.