0

I am studying machine learning, and I am very confused about the definition of "prediction" and "label", and I would like to know what is the relationship between them?

My understand is: "prediction" is something you are going to predict, based on "label".

E.g., label = MCQ1 MCQ2, prediction = Final_term_mark

it can predict a student's final term mark by his/her grade of MCQ1 and MCQ2.

Is that correct?

Has QUIT--Anony-Mousse
  • 76,138
  • 12
  • 138
  • 194
seki
  • 89
  • 1
  • 7
  • What are the definitions that you know for prediction and label? – SBylemans Oct 30 '18 at 12:19
  • @SBylemans Hi, thanks for your reply, I just update the description of this question. – seki Oct 30 '18 at 12:37
  • Possible duplicate of [What is the difference between a feature and a label?](https://stackoverflow.com/questions/40898019/what-is-the-difference-between-a-feature-and-a-label) – Neo Oct 31 '18 at 14:48
  • I think prediction comes later. First labels along with features are used to derive information and then predictions can be done based on the given scenarios. – Darshan Jain Feb 13 '20 at 09:56

2 Answers2

1

Labels are the known values for old data.

Prediction is your predicted value for new data, where you do not have a label (or pretend that you do not have a label - in evaluation).

During training, you try to make your predictions match the labels.

Has QUIT--Anony-Mousse
  • 76,138
  • 12
  • 138
  • 194
  • No, you train your model so its predictions match the labels. So you create predictions for both *old* (i.e. annotated/labeled) and *new* (unlabeled) data. – cheersmate Oct 31 '18 at 10:09
  • You do that also, so it doesn't contradict my statement. But I will add that more obvious use of the labels. – Has QUIT--Anony-Mousse Oct 31 '18 at 13:27
  • The machine learning model is trained on the label length. The label length refers to the number of timesteps in the future for which the ground truth values are available during training. In other words, the model is trained to predict the values of the time series at time steps that fall within the label length given the values of the previous time steps. The prediction length, on the other hand, specifies the number of timesteps in the future for which the model is asked to make predictions during inference after the training is completed. – freak11 Feb 18 '23 at 15:26
0

According to developers.google.com. A label is a thing we're predicting—the y variable in simple linear regression. The label could be the future price of wheat, the kind of animal shown in a picture, the meaning of an audio clip, or just about anything.

Predictors are more like an estimator that is capable of making prediction like a Linear Regression model.

Manan Jain
  • 27
  • 5