African Credit Scoring Challenge
Helping Africa
$5 000 USD
Completed (almost 2 years ago)
2064 joined
1020 active
Start
29 Nov 24
Close
12 Jan 25
Reveal
13 Jan 25
does the economic_indicators help
Help · 29 Dec 2024, 17:01 · 12
i still didn't add them to my data but does they improve my model and any tips on how to add them
Moi j ai essayé de les ajouter mais il sont sans effet positif. Pour ce faire j ai juste merger les deux bases.
Ce que vous dites est vrai. J'ai essayé la même chose et j'ai obtenu un score f1 de 0,0 (ce qui signifie qu'il prédit uniquement la classe 0). Je ne dirai pas de ne pas l'utiliser, mais il pourrait y avoir un moyen d'obtenir un score de descente. Essayez "class_weights" si vous utilisez GBT.
They did for me. You just have to figure out how best to use them.
Did you use class weights or tried to balance the classes?
well i have some ideas to use it but it is a little annoying to code so i'm running away from it :)
You have to try stuff out to see what gives you the best results
class balancing doesn't seem to work right now, haven't tried class weight yet
Well to me not yet. Not yet because I tried but it didn't improve the model. However you can still try it out to see if it works out. Try using class weights if you are getting an f1 score of 0
I tried and it didn't give me any noticeable improvement. I'm sure I am not using them rightly though.
I merged it with the tain and test data sets considering the specific factors for the two countries Kenya, Ghana, but it drastically impacted the accuracy, it seems the econmic indicators need more investigations to make the direct relation between these factors and loan default prediction.
It did for me :) you just have to figure out the right combinations
i'm doing it right now chatgpt made me soft so i'm going to look into that