JamesonMemorialBets
Predicting Elite: Dangerous market prices with the power of machine learning™
Data sources
Live market data from EDDN Mass data dumps from EDDB Mass data dumps from EDSM
Reference material
Elite: Dangerous player journal manual
Usage
Basic input data (such as that generated by collect_market_data.py) is included in the repository using git LFS.
However, it still needs to be merged and cleaned before predictions can occur. Once the repository and data are
downloaded, run the following commands to process the data and make predictions.
Important note! Performing predictions with the provided data set is very memory intensive, using approximately 8.5 GiB of memory on our systems.
python merge_data.py
python clean.py
python predict.py
By default, predict.py will use the pre-trained module in the repository.
You can delete this model (bets.tf) to instead train a new model.
If you want to use a different set of data market/event data, run collect_market_data for a period of time to collect
live events from players. If you want to use a different set of data for static system metadata, download the
"Populated Systems" data dump from EDSM and save it as
data/systems_populated.json.gz, then run parse_system_data.py. You'll need to merge and clean the data again if you
do either of these.