Concurrent_Neural_Network

Purpose & general Presentation

This code represent supplementary material for the paper Concurrent neural network: a model of competition between times series ( http://link.springer.com/article/10.1007/s10479-021-04253-3 ), published in Annals of Operations Research. To sum up, it present a embedding of a neural network f. This embedding rescale its output to a given neural network to match a given sum. The i-th component of its output is then :

$$y_i = \frac{f_i(x)}{\sum_{j} f_j(x)}$$

This allows us to train this neural network for predicting market share of competing assets.

Requirements

This code was tested using Python 3.7. The following packages were used for the core package

• numpy==1.18.5
• torch==1.8.1
• pandas

sklearn was used to provide benchmark.

Organisation of the code

• Concurrent_Neural_Network : contains the main source code for the project
  • models.py : definition of the main Concurrent_Module module presented in the article
  • submodel.py : definition of the subneural network Multi_layer_feed_forward_model
  • preprocessing.py : various tools used to preprocessed data
• example_notebook.ipynb : Jupyter notebook showing a simple example of use of the module
• example : outdated example of use of preprocessing tools
• data : A small sample of data used in example