2019-04-16 01:18:09,712 INFO Model: all_days-5.msgs-30-words-40_word_embed-glove.vmd_in-hedge_alpha-0.5.anneal-0.005.rec-zh_batch-32.opt-adam.lr-0.001-drop-0.3-cell-gru, session restored!
2019-04-16 01:26:11,191 INFO Eval, eval loss: 1.03172838688, acc: 0.574405This repository releases the code for stock movement prediction from tweets and historical stock prices. Please cite the following paper [bib] if you use this code,
Yumo Xu and Shay B. Cohen. 2018. Stock Movement Prediction from Tweets and Historical Prices. In Proceedings of the 56st Annual Meeting of the Association for Computational Linguistics. Melbourne, Australia, volume 1.
Stock movement prediction is a challenging problem: the market is highly stochastic, and we make temporally-dependent predictions from chaotic data. We treat these three complexities and present a novel deep generative model jointly exploiting text and price signals for this task. Unlike the case with discriminative or topic modeling, our model introduces recurrent, continuous latent variables for a better treatment of stochasticity, and uses neural variational inference to address the intractable posterior inference. We also provide a hybrid objective with temporal auxiliary to flexibly capture predictive dependencies. We demonstrate the state-of-the-art performance of our proposed model on a new stock movement prediction dataset which we collected.
Should you have any query please contact me at yumo.xu@ed.ac.uk.
- Python 2.7.11
- Tensorflow 1.4.0
- Scipy 1.0.0
- NLTK 3.2.5
- src: source files;
- res: resource files including,
- data: datasets consisting of tweets and prices which you could download here.
Model configurations are listed in config.yml where you could set variant_type to hedge, tech, fund or discriminative to get four corresponding model variants, HedgeFundAnalyst, TechincalAnalyst, FundamentalAnalyst or DiscriminativeAnalyst described in the paper.
Additionally, when you set variant_type=hedge, alpha=0, you would acquire IndependentAnalyst without any auxiliary effects.
After configuration, use sh src/run.sh in your terminal to start model learning and test the model after the training is completed. If you would like to do them separately, simply comment out exe.train_and_dev() or exe.restore_and_test() in Main.py.