ANN model of it is a simple, the front is a few influence factors (ID, college, degree, gender, gpa), the last is a model to predict - income

In college I used normal encoding, degree and gender is one hot encoding

Before using the dichotomy to replace with 0 s and 1 s, online writing a number of classification can be in this way:

I think that to classify, 0-15000 baht for 0150 01-30000 baht of 0.5, 30001-200000 baht to 1, three classification, finally come out of the data at this range (0, 1), and then classified as 0 0-0.33, 0.33-0.67 the classification of 0.5, 0.67-1 to 1. In this way to determine the precise degree model, this is my code, the data preprocessing in front of the no problem, and finally a number of classification, I won't answer,









PS:

Part 4 parts of predicting the test set the result is a problem, I don't know the loss from binary_crossentropy changed to categorical_crossentropy correctly or not