Different predictions made by R code and R predictor node

[dmitch52]

Hi KNIME community, 

Can anyone think of why predictions made from R code in R are different from predictions made using the R learner/ R predictor node in KNIME with the same R code?

Thanks

[mauuuuu5]

Well I was having the same issue "predicting" a cluster in Knime, and I looked at the internet and found that you should use set.seed() beforehand. 

See this link: https://stats.stackexchange.com/questions/26364/variability-in-k-means-clusters-results-setting-set-seed-before

hope it helps

[johannes_clarifydata]

This (use set.seed() for reproducibility) also applies to various other prediction algorithms (e.g. tree-based algorithms). Standard OLS should lead to the same results, though. ;-)