Good day support team, I am having some difficulties implementing the automatic mlforecast. I try to apply the code to the attached dataset but get several errors. Specifically: I got errors of: "AutoRandomForest: ValueError: Some value(s) of y are negative which is not allowed for Poisson regression. AutoLinearRegression: ValueError: Found input variables with inconsistent numbers of samples: [134, 12] ValueError: Found input variables with inconsistent numbers of samples: [134, 12]"

Hey. Thanks for the thorough report. The error for random forest is coming from the fact that we're applying differences (which produce negative values) and trying a poisson loss. We'll remove the poisson loss, in the meantime you can manually remove it by defining a different space like this:

from mlforecast.auto import AutoModel, random_forest_space
from sklearn.ensemble import RandomForestRegressor

def my_rf_space(trial):
    return {k: v for k, v in random_forest_space(trial).items() if k != "criterion"}

my_auto_rf = AutoModel(RandomForestRegressor(), my_rf_space)

The other errors are coming from your custom loss. The training dataframe is provided in case you want to compute a loss that uses the training set, such as MASE. In this case you're comparing the forecasts vs the training set values of the target. I believe what you want to do is this instead:

def mse_loss(forecast_df, train_df):
    y_true = forecast_df['y'].values
    y_pred = forecast_df['model'].values
    return mean_squared_error(y_true, y_pred)

Thank you so much Jose. All worked out for now. Much appreciate it ^_^