[Predictive intelligence] Why is the Regression solution giving negative (-ve) Lower Bound values when the Confidence Level is set to 95%, even though there are no negative values in the training data.Summary Question: Having trained a Regression solution to predict the incident "Resolve time (mins)", when testing the solution via the Test Solution tab, when using 95% confidence values, the LowerBound value is returned as a negative value, indicating that the incident is resolved before it was raised. However, in the Regression solution training data, all the values for "Resolve time (mins)" are positive. The LowerBound only returns a positive value when the confidence level is set to a much lower level. Why? Answer: This behaviour with negative LowerBound values in a Regression solution is expected. Based on the training data, we calculate the variance of "actual vs predicted" values on the training data, and use that to provide the bounds based on confidence level. The higher the confidence levels, the larger the range (i.e. bigger difference between the lower bound and upper bound values). In a way, this denotes that the actual value will be in the given range with the selected confidence level. Depending on the range of the training data, for predictions that return lower values where the confidence level is set high, the case arises, where the lower bound will be shown as a negative value. However, the predicted value itself will never be negative. Question: So what is the range of acceptable confidence levels, if the LowerBound only returns a positive value, when the confidence level is set to a much lower level than 95%? Answer: There is no "ideal" confidence level, as each trained model is different. The model returns predictions and you can adjust the confidence level as per your requirement for this Regression solution. The higher the confidence level, the bigger the difference between the lower bound and upper bound values, in which it is confident that the predicted resolve time will be positive and within these Lower/Upperbound values. Hence, when you lower the confidence, the difference between the lower bound and upper bound values get smaller, and so it will be less confident that the predicted time will actually be within this smaller range.