Hi, I am trying to tune the parameters of a classification problem. Can I use MSE as fitness function in PSO or is it only for regression problems? If yes, what fitness function can be used for classification problems?
Thanks
How do you handle MSE as a metric for classification? Why don't you use metrics such as F1-score for classification tasks? If you want to adjust the hyperparameters of a classification method/model, you should define the cost-function based on the performance of the classification, such as using metrics like F1-score or overal accuracy. This will help to optimize the hyperparameters and improve the model's performance.
Hi, I tried using F1-score, accuracy, recall as well however, it is not working well with the model. Can hamming loss be used instead?
Hi Akshita Dhiman , Sure, you can use Hamming loss. It measures the fraction of labels that are incorrectly predicted. I'm puzzled why you didn't get good results by using F1? Did you directly use F1 as the metric? Because you should use 1-F1 or 1/F1.
Hi Arvin Fakhri Thank you for your response. I guess I was using the metric directly. Also, I tried using 1-f1_score and 1/f1 however, I am getting same results every time. I have tried increasing the no. of particles or iterations, but the results are still the same.
- Badges
- Science topic
- Similar topics
- Indexes