How to fit integer output in neural network

Question

I have a data set that contains 104 input features and 96 output values ​​to be predicted, the input features are floating values ​​normalized to [-1,1], and each out put value would be one integer among [-1,0,1].

I am basically new to data science and machine learning, and I need to know what kind of neural network model would fit best for this kind of data. I have googled this for a quite long time, can you guys help me ? Thank you!

我的每一条训练数据具有104个[-1,1]之间的浮点型输入量和96个整数型输出量，每个输出值为[-1,0,1]之中的一个整数。
我是机器学习的新手，我想知道应该采用何种神经网络模型来解决此问题，在post本问题之前我已经自行搜索了很久，希望大家能给我一些帮助，谢谢！

Answer 1

You may check a multitarget model like this one. Basically Keras‘ functional API gives you a lot of flexibility to model multitarget problems.

You have three classes per target, so make sure you use 'categorical_crossentropy' as loss function with three terminal layers.

Answer 2

If the output range is just three integers (-1, 0 and 1), the problem would probably be best approached as multinomial classification. For a start, you can look up a few multinomial classification tutorials for scikit-learn for Python: https://scikit-learn.org/stable/index.html.