Project created for "Data mining concepts and techniques" course.
Use the zoo dataset to design a neural network based classification system to predict the species of an animal. You should use a K-fold validation strategy to judge the overall accuracy of your system. Does the accuracy of the system change when you vary the number of folds to use.
I’ve been able to make a neural network based classification system to predict the species of an animal with an accuracy score of 97.14% using a K=5 fold validation strategy to judge the overall accuracy of my system.
Yes, the accuracy of the system varies when we change the number of folds. It increases till K=5 and then decreases and stabilizes around 95.00-96.00% even if we increase the epochs parameter. The accuracy varies as such with K (No. of folds):
|No. of folds||Accuracy|
I’ve used ‘type’ attribute as label, Z-transformation method to normalize the data, a deep learning model with epochs = 1000. I’ve used accuracy and classification error as my performance metrics.