stockAI is an integrated machine learning package for technical research of stock price big data. From data collection to machine learning-specific stock price time series dataset generation, AI model learning and model evaluation, and backtesting, the goal of this package is to provide a one-time Python library to enhance the utilization and convenience of stock price big data technical research.
- Stock Price Big Data Collection
'stockAI' stores big data in github and provides it to users. Because 'stockAI' is a research package, the data provided here focused on collecting data, storing it reliably, and storing it in a high-quality way, rather than collecting it in real time with unstable crawling methods.
- An integrated package that enables the entire process of machine learning
In the past, we had to use distributed libraries to study stock price big data. StockAI is very useful because it is an integrated package that can collect data, preprocess data, learn artificial intelligence models, evaluate models, and backtest (yield calculation) at once.
- Support the convenience of stock price big data research
The stockAI package can also be used by experts with domain knowledge of stocks. When these experts are unfamiliar with programming and want to perform artificial intelligence technological analysis with stock price data, it helps them conveniently conduct research on stock price data without a large amount of programming.
You can find more information about using stockAI in the tutorials folder.
And The following image summarizes the methods of stockAI.
1.1 Bringing up the market
sai.get_markets(country:list)1.2 Bringing up tickers
sai.get_tickers(markets:list, date:string)1.3 Bring up stock price data
sai.load_data(date:list, tickers:list=None)2.1 Add Secondary Indicators
sai.add_index(data:pd.DataFrame(), index_list:list)2.2 Scaling
sai.scaling(data:pd.DataFrame(), scaler_type:String, window_size:Int=None)2.3 Convert to time series data
sai.time_series(data:pd.DataFrame(), day:Int=10)3.1 trader definitions
For trader definitions, it is recommended to refer to the tutorials folder.
3.2 Store required data sets inside the trader
sai.save_dataset(lst_trader:list, train_data:pd.DataFrame(), test_data:pd.DataFrame(), train_data_scaled:pd.DataFrame()=None, test_data_scaled:pd.DataFrame()=None)3.3 Model fitting
sai.trader_train(lst_trader:list)3.4 Model evaluation
sai.get_eval_by_threshold(lst_trader)3.5 Setting Thresholds
sai.set_threshold(lst_trader, lst_threshold:list, hisogram:bool=True)4.1 Making a sales log
sai.decision(lst_trader:list, dtype='test', data=None, data_scaled=None)4.2 Calculate the yield
sai.simulation(df_signal_all, init_budget, init_stock)4.3 Leader Board
sai.leaderboard(df)4.4 Visualize yield
sai.yield_plot(df)dfsf