LFM Quant

by Euclidean Technologies, LLC

On a periodic basis, publicly traded companies are required to report fundamentals: financial data such as revenue, operating income, debt, among others. These data points provide some insight into the financial health of a company.

This repository contains a set of deep learning tools for forecasting company future fundamentals from historical fundamentals and other auxiliary data such as historical prices and macro economic data.

Installation and Setup

Clone repo, setup environment, and install requirements:

$ git clone https://github.com/lakshaykc/lfm-quant.git
$ cd lfm-quant
$ export LFM_QUANT_ROOT=`pwd`

You may also want to put LFM_QUANT_ROOT in your shell initialization file such .bash_profile so that it does not need to be defined every time you start a shell. For example, you could run the following from within the lfm-quant directory:

$ echo "export LFM_QUANT_ROOT="`pwd` >> ~/.bash_profile

Building Models

You can train lfm quant on a neural network of a particular type and of a particular architecture with several other hyperparameters on a particular dataset by first defining all of these things on a config file, and then specifying that config file as the point of reference when running lfm_quant.py. Consider, for example, how lfm_quant is run on open-dataset.dat, as specified by config/system-test.conf:

$ python scripts/lfm_quant.py --config=config/system-test.conf --train=True

This will load the corresponding data file. Model checkpoints are saved in a directory defined by model_dir parameter in th config file. A couple of notes about config files:

• The user can specify a .dat file to use through the --datafile and the data_dir options (note that the latter is datasets by default).
• financial_fields is a range of columns, and should be specified as a string joining the first and last columns of the .dat file that the user wants to forecast (for example: saleq_ttm-ltq_mrq).
• aux_fields is similarly also a range of columns that is equivalently specified. Note, however, that these fields are strictly features; they are not part of what the model is trained to predict.

Generating Forecasts

To generate forecasts for the companies in the validation set, lfm_quant.py must be run with the --train option set to False. For example:

$ python scripts/lfm_quant.py --config=config/system-test.conf --train=False

This uses checkpoint directory created under model_dir for inference and writes resulting files in the same dir.

Dataset

Paper [ref] uses data licensed from Compustat North America. The dataset is proprietary and cannot be made public. We provide notes in dataset_api.md to create a module dataset. This module converts data from source to tensorflow.datatset object type.

Contributors and Acknowledgement

This repository was developed and is maintained by Euclidean Technologies, LLC. Individual core contributors include Lakshay Chauhan, John Alberg and Zachary Lipton

License

This is experimental software. It is provided under the MIT license, so you can do with it whatever you wish except hold the authors responsible if it does something you don't like.