Home Forums Main Forums Big Data Lift curve in decile analysis

  • Lift curve in decile analysis

     Justin updated 1 year, 1 month ago 1 Member · 1 Post
  • Justin

    Administrator
    October 28, 2020 at 10:05 pm

    Decile analysis is widely used to assess the performance of a predictive model. Below is one decile table.

    However, different people use different definitions for lift ratios. Some people just define it as:

    Lift= Response rate of a decile/ the overall response rate

    Use above table, the lift will be 12.30%/4.20%, 7.30%/4.20%, 5.90%/4.20%….. 0.30%/4.20%. In this case, the lift value will decrease to a limit of zero.

    On the contrary, some people prefer to use cumulative response rate to define it:

    Lift= Cumulative response to a decile/ the overall response rate

    Then the lift will be 12.30%/4.20%, 9.80%/4.20%, 8.50%/4.20%….. 4.20%/4.20%. The lift value will decrease to 1 eventually. Above table uses this cumulative lift definition. Please note: the cumulative lift values were multiplied by 100 actually in above table.

    Either one is good to use, it’s up to you to choose.

Log in to reply.

Original Post
0 of 0 posts June 2018
Now