We build a neural network which have one hidden layer. And got good accuracy for training and test set. This code-section was help to visualize basic structure of nn. There have a lot of work to improve.
- Dataset: Churn Modealing (very known business dataset)
- Cost function: MSE
- Activation: sigmoid (can use relu!)
- accuracy: arround 80% for both train and test dataset
- Used modeule: numpy, pandas, matplotlib and sklearn (for preprocesing and model evalution checking). For learn more about those modeule go here.