Transform data to be digestable by scikit-learn

transform_sgd_data.py

TRADING_DAYS_IN_2019 = 252
TODAY_NUMERIC_VALUE = 0
from datetime import datetime
end = datetime.strptime('2019-02-07', '%Y-%m-%d')
shape = 40000
train_valid_tickers = 2392
test_valid_tickers = 800

def transform_data(test_or_train_set, valid_tickers):
    y_set = []
    x_set = np.zeros((valid_tickers,shape))
    index = 0
    ticker_map = []
    for ticker in test_or_train_set:
        if len(ticker[0]) > TRADING_DAYS_IN_2019 + 1:
            y_set.append([ticker[0]['5. adjusted close'].to_numpy()[TODAY_NUMERIC_VALUE] > ticker[0]['5. adjusted close'].to_numpy()[TRADING_DAYS_IN_2019]])
            values = ticker[0].loc[ticker[0].index < end].values.flatten()
            internal_index = 0
            for value in values:
                x_set[index][internal_index] = value
                internal_index = internal_index + 1
            index = index + 1
            ticker_map += [ticker[1]['2. Symbol']]
    return x_set, np.asarray(y_set).reshape((valid_tickers,)), ticker_map
    
x_train_set, y_train_set, train_ticker_map = transform_data(train_set, train_valid_tickers)
x_test_set, y_test_set, test_ticker_map = transform_data(test_set, test_valid_tickers)