I trained data set and tested (few no of data) using SVM, Multiclass SVM and ANN in MatLab. But I need to know the performance accuracy of these classifiers. Is there any functions are available in MatLab? otherwise suggest me any techniques to get accuracy value (other than precision, recall, F measure)
This a good question indeed, we have to think about two terms performance and accuracy.
1. A bulky neural network which require high computation time can be an accurate one but what about the computation burden?
2. An optimal neural network which is designed with less no. of hidden layers and constructed through feature selection routine can be less accurate but it put less computation burden and swift enough.
a few years back we did this study with the help of design of experiments to design neural network at micro and macro level state. the same goes with the SVM, for kernel selection and bias parameter estimation.
My fellow RG friends and colleagues have atached many papers for your reference.
Regards
If you train your model with Matlab Classification learner it automatically makes confusion matrix for validation data accuracy, and for test data you need to count the number of true positive false positive true negative false negative to calculate Precision and recall score.
if you are not familiar with confusion matrix follow link
https://en.wikipedia.org/wiki/Precision_and_recall
follow link.
https://kr.mathworks.com/help/stats/classificationlearner-app.html
This a good question indeed, we have to think about two terms performance and accuracy.
1. A bulky neural network which require high computation time can be an accurate one but what about the computation burden?
2. An optimal neural network which is designed with less no. of hidden layers and constructed through feature selection routine can be less accurate but it put less computation burden and swift enough.
a few years back we did this study with the help of design of experiments to design neural network at micro and macro level state. the same goes with the SVM, for kernel selection and bias parameter estimation.
My fellow RG friends and colleagues have atached many papers for your reference.
Regards
Confusion matrix. And Classification learner in MATLAB can help you out in deciding best classifier for your data
- Badges
- Science topic
- Similar topics
- Organizational Psychology
- Training