Reinforcement Learning for Capstone data that does not exist on first run

AI & Machine learning
1

The Capstone Model solution requires data for WEIGHTS_FILE, REPLAY_FILE and TRADE_FILE. Assuming that the TRADE_FILE is the sample data used in Part II, SPY SP&500. Then WEIGHTS_FILE and REPLAY_FILE, does not exist on a first run.



If you try to do a first run with only the raw data with SPY, as the 'TRADE_FILE':  'dataframes = [pd.read_csv(f) for f in filenames]', then there is no processing of the Capstone Model Solution.ipynb.



If I replace the dataframe with PriceData5m.bz2, then again there is no processing of the Captsone Model Solution.ipynb.



What is the procedure for doing a first run of the Trade_File to generate the Weights_File and the Replay_File?



Please advise,



William Porter

 

2

Hi William,



I think there is a bit of confusion. The 'TRADE_FILE' is used to store the trade level data when the deep reinforcement model is run on the input data. The input data used as an example is the FX pair data stored in the file 'fx_pair_data_1.bz2'.



Due to the resource intensiveness of the deep learning algorithm, it is recommended to download the notebook and related files to your local system and then run the notebook on the local system itself. You can download the capstone-related files from section 22 unit 7 of the course. 



You can run the deep reinforcement algorithm on the "PriceData5m.bz2", file as well, provided you specify the correct path in your code. The course-related files are present as a downloadable zip file in the last section of the course, ie section 25 unit 2.



Hope this helps.