I can't put glass.csv dataset because of copyright rule. Please see in UCI Machine Learning Repository. Please help me to get this assignment. Thank you.
File name glass.data is in the CSV format.
Train your fully-connected neural network to classify these glass according to this guideline.
1. This is a multiclass problem. The target output should be converted to one-hot of six types.
2. Classes are not balanced. You should perform class balancing.
3. Perform data normalization.
Print and show the graphs of
1. Loss of the training set
2. Accuracy of the training set
during the progress of training in each epoch.
Please note that your network and parameters should achieve a reasonable loss and accuracy. You have to try and find your own parameters. Use techniques that were mentioned in class to achieve a good result. You should try to normalize your data and perform class balancing (you can choose to do these by hand in Excel or by programming in Python).
The glass piece left in the crime scene has these features
(1.52, 12.8, 1.6, 2.17, 72.2, 0.76, 9.7, 0.24, 0.5)
After training, put this data into your model and see what type that your model classifies to. Show the code + print out result of this part and give the answer in the report. Also show the probability value from the softmax of this class, which indicates the confidence of your answer.
