ptyagi / CryptoSignal / 1.0.0


This program uses LSTM neural network to predict the trend of any ERC20 token. It extracts realtime data from Santiment, a dApp on Ethereum.

For data collection module please refer to

Applicable Scenarios and Problems

What scenarios or problems would this algorithm work well in?


Although one can choose finer intervals like 5minute, 10minute etc but it is expected to work better for day data or hourly data.

The latest data is imported on the fly and model is retrained before doing prediction for next sample.


As input it requires to provide the token name, days, and interval like ["tokenname",days,"interval"] List of token names is given below:

"ethereum","binance-coin","omisego","augur","maker","mithril","kucoin-shares","wax","bancor","loopring","decentraland","digixdao","dentacoin","crypto-com","funfair","storj","sonm","sirin-labs-token", "mobilego","huobi-token","indorse-token","aragon","power-ledger","tenx","kyber-network","salt","civic","singularitynet","santiment","singulardtv","0x","aeternity","basic-attention-token","bytom","golem-network-tokens","chainlink","populous","status","waltonchain","aion" etc


Once you give the input to the program and run it -it will output the trend for the token like "No Change", "Down", "Up"


input = ["ethereum",20,"1d"] Output= "Up"