Hi, I want to predict the salary of people working in financial market and I know their age, health-insurance level and marital condition and so on. However, the sample that have a salary above 250000$ per year is not ample enough compared with those who have a salary below. So the performance of the high salary is very poor, although the performance on the whole is good. Can you provide me some method to improve the the accuracy of the high salary? Thank you very much!
Hello Chen!
There are several methods to improve the performance on machine-learning algorithms for imbalanced classes. Unfortunately there is no one algorithm that can work correctly for each method and dataset as they all have pros and cons (no free lunch).
I am attaching a presentation that I created some time ago. This presentation contains some of the methods that are used most often. They are fairly easy to implement (specially the sampling ones) so it would be a good idea to test several of them to see which one gives better results.
I hope it helps you with your application!
Have a nice day!
Hello Chen!
There are several methods to improve the performance on machine-learning algorithms for imbalanced classes. Unfortunately there is no one algorithm that can work correctly for each method and dataset as they all have pros and cons (no free lunch).
I am attaching a presentation that I created some time ago. This presentation contains some of the methods that are used most often. They are fairly easy to implement (specially the sampling ones) so it would be a good idea to test several of them to see which one gives better results.
I hope it helps you with your application!
Have a nice day!
Thank you for your presentation! There are several useful method to cope with the imbalanced data, I will learn them gradually.
This post can give you some pointers
http://machinelearningmastery.com/tactics-to-combat-imbalanced-classes-in-your-machine-learning-dataset/
- Badges
- Science topic
- Similar topics
- Public Health
- Health