Please check this dataset at the UCI Machine Learning repository http://archive.ics.uci.edu/ml/datasets/Credit+Approval First, read the description, and then download the “crx.data”, a CSV file.
- Build a prediction system based on Artificial Neural Network (ANN) that can predict a person’s credit-worthiness (+ or -).
- Please try various architectures of the ANNs and report test evaluation results (Accuracy, Precision, Recall, and F1).
- Each of the architectures should vary the following:
- Number of layers
- Number of neurons (i.e., perceptrons) per layer
- Activation functions (e.g., sigmoid, RELU, etc.)
- Using Kera’s model.summary() function, print each of the architecture.
- Try different hyper-parameters for each of the above architectures:
- Number of epochs: 100+
- Learning rate: 0.01+
- Please list potential ethical bias inherent to solving the problem. And discuss how are you going to address the issue.