New answers tagged

0

Your "adviser" can use the correlation between the explanatory variables and the explained variable. You can also use information provided by the p-value More details here : https://towardsdatascience.com/feature-selection-correlation-and-p-value-da8921bfb3cf


1

Since you partly overfit with RF, first try to get the RF hyperparameter right. You could do a grid search like: rf = RandomForestClassifier(...) param_grid = { 'n_estimators': [200,300], 'max_features': [10,20,30] } cv = GridSearchCV(estimator=rf, param_grid=param_grid, cv= 5) cv.fit(xtrain, ytrain) In RandomForestClassifier max_depth and ...


1

You could try a RegressorChain() to predict $\hat{y}_1,... ,\hat{y}_n$ and compare the results to a model where you predict $\hat{y}_{total}$. RegressorChain() may help to pick up useful information regarding the $y_1,...,y_n$ in the modeling process: Each model makes a prediction in the order specified by the chain using all of the available features ...


0

For machine learning frameworks, it is important that the inputs are within well-specified bounds. Requiring target labels to be positive makes subsequent code and interpretation much easier.


Top 50 recent answers are included