What does a neural network look like?

Question

Neural networks

The task is to write a neural network that makes a decision on issuing a loan.
Guys, tell me what the point is, do not kick.
If you make a database with questions and weights, questions and a sign, where to go next - is this a decision tree?
How is a neuron different? What does a neural network database look like? Guys, I ask you to show practical examples, the insides of the neural network and the code, I reviewed a whole bunch of books, they didn’t help much.
Many thanks to everyone who replies :)

Answer 1

Google decision trees( tyk , tyk ), there is a classic example about the problem of credit scoring.
According to neural networks, based on the keras framework, it looks something like this:

from keras.models import Sequential
from keras.layers import Dense, Activation

model = Sequential()
# Полносвязный слой из 32 нейронов, принимает на вход вектор из 25 признаков
model.add(Dense(32, input_shape=(25,)))
model.add(Dense(64)) # Полносвязный слой из 64 нейронов
model.add(Dense(1)) # Выходной слой
model.add(Activation('relu')) # Полулинейная функция активации

# Оптимизатор: Root Mean Squared Error Propagation
# Функция потерь: Logloss aka перекрестная энтропия
# Метрика: accuracy
model.compile(optimizer='rmsprop',
              loss='binary_crossentropy',
              metrics=['accuracy'])
...
model.fit(Xtrain, Ytrain)
accuracy = your_eval_func(model.predict(Ytest, Xtest))
preds = model.predict(target_data)

Answer 2

You can install this https://github.com/Microsoft/LightGBM
It has installation instructions, documentation and usage examples.