Mean Absolute Error increasing with more correlated factors

External question — Data Science Stack Exchange Author: user68978 Original post: https://datascience.stackexchange.com/questions/46762 License: CC BY-SA 4.0 — https://creativecommons.org/licenses/by-sa/4.0/ Adaptation: HTML converted to plain text; contact email addresses removed. I am using Microsoft Azure Machine Learning Studio to predict stock market prices. We have the variables- Index price(target-to be predicted),Low price,High price,dates and days. We use split of 0.7 and run Linear regression. We get Mean absolute error of 109. We then try to add more variables(macroeconomic factors which positively effect the index prices) which are correlated with the target variable and should improve the predictions- we find that the Mean Absolute error increases to 110.I have attached the pics for your reference. Are we interpreting wrong or what's wrong in we are doing? PS:We tried Boosted Tree regression as well-but the same problem as described above is observed. Errors

External answer — Data Science Stack Exchange Author: Sunil Original post: https://datascience.stackexchange.com/a/46764 License: CC BY-SA 4.0 — https://creativecommons.org/licenses/by-sa/4.0/ Adaptation: HTML converted to plain text; contact email addresses removed. Any reason why you are using MAE instead of MSE? The reason for using MSE is that you get a parabola which would be more accurate for the optimizing algorithm to find the local minima Did you check the relation between predictors and target variable? did you check multi collinearity?