spss multi-indicator ROC curve prediction

lidongmei_ 注册会员
2023-01-25 18:12

According to the problem you described, if the 3 separate indicators are above the roc diagonal, then they change after combination. Then you need :

  1. Check whether the method of joint index analysis is correct and whether there are any mistakes in the middle. Check and check step by step. Generally, joint index analysis takes multiple indicators as logistic regression to get joint factor, and then carries out roc analysis with joint factor as the new index.
  2. roc curve should be below the curve. The roc curve you draw curves downward, which may be because the coding level of the dependent variable is reversed. The solution is to adjust the direction. You can adjust the direction directly in the options of spss software to ensure that the roc curve meets your requirements.
    Answer is not easy, rich to give a reward, no money to an adoption, passing by to give a praise, thank you!!
dats7792 注册会员
2023-01-25 18:12

How do I get both the combined and the separate curves above the guide line?

dmq0708 注册会员
2023-01-25 18:12

Please accept it!
In the ROC curve, if the curve is above the reference line, the performance of the classifier is good; if it is below the reference line, the performance is poor.

When multiple indicators are combined, the curve is found below the reference line. This is because the performance of the joint indicator may be worse than that of a single indicator, or the weight of the joint indicator is inappropriate.

Solutions may be as follows :

1. Improve performance by adjusting weights
2. Use another type of classifier
3. Filter effective features by feature selection