CODEWITHSUNDEEP
pythonmachine-learningregression
jiro.kokubunji
jiro.kokubunji

Reputation: 51

Way to predict with regression model

I have a question about regression model in machine learning and I am wondering if my way is correct or not.

I have built my regression model and already trained it with my data, but my model always predict 1.5 times more than actual values.

I understood that this is my model's habit, consider as it is predict alway 1.5 times.

After considering as it is, I divided predicted value by 1.5 times. Let's say, my model predict 100 in some case, and I calculated 100/1.5 and get approximately 66.6 in a result.

Actually 66.6 is not predicted value and I manipulated it.

Is this manipulation acceptable for regression? Can I supply this 66.6 to my customer?

Upvotes: 0

Views: 92

Answers (1)

Dat Tran
Dat Tran

Reputation: 2392

No you shouldn't do it. If your mode always predict 1.5 times more than the actual values then that means your model is just not performing well and the data cannot be linearly fitted. To prevent this, you should look at other models that is able to capture the structure of your data or you might have outliers and removing them would help the linear regression model.

Upvotes: 1

Related Questions