search

lilianweng / stock-rnn

Predict stock market prices using RNN model with multilayer LSTM cells + optional multi-stock embeddings.
https://lilianweng.github.io/lil-log
1.72k stars 656 forks source link

some changes #21

Open megazone87 opened 6 years ago

megazone87 commented 6 years ago
  1. py2 to py3
  2. As google finance data is unstable, especially in china, "pandas_datareader" and its "quandl" api is used for data_fetcher
  3. data batching is changed:
    • previous: target of current sample is the last value of next sample in the same batch.
    • now: make taget seprate from input, by shuffle the samples in whole set, thus they don’t related within one batch.
  4. sort price by date from old to new date, thus the prediction is about recent date.
  5. pandas deprecated: sort => sort_value

Thank you for checking!