PropensityToBuy

Model built using sklearn which predicts probability of purchasing financial given input the bank transactions of users

Different Approachs Used

    
    * Data type check
    * Importance variable selection
    * Standardize the continuous variable(scale)
    * Neural network
    * Ada boost

Mutual Information is used to select Features

Neural Network model architecture:

Charecteristic of the NN model:

• Optimizer : Stochastic Gradient Descent(For training Examples > 1000 it converges faster)
• Activation function : Logistic (sigmoid).
• Cost Function : Logistic Cost Function.
• Learning rate : Adaptive.
• Features selection : Select 50 best using mutual information.

AdaBoost model architecture:

Charecteristic of the AdaBoost model:

• Optimizer : Stochastic Gradient Descent (For training examples > 1000 it converges faster).
• Cost Function : Logistic Cost Function.
• Features selection : Select 50 best using mutual information.
• Base Classifier: Decision Trees.