Function cal_factor

finagent/processor/processor.py:61–195

(df, level="day")

def cal_factor(df, level="day"):
    # intermediate values
    df['max_oc'] = df[["open", "close"]].max(axis=1)
    df['min_oc'] = df[["open", "close"]].min(axis=1)
    df["kmid"] = (df["close"] - df["open"]) / df["close"]
    df['kmid2'] = (df['close'] - df['open']) / (df['high'] - df['low'] + 1e-12)
    df["klen"] = (df["high"] - df["low"]) / df["open"]
    df['kup'] = (df['high'] - df['max_oc']) / df['open']
    df['kup2'] = (df['high'] - df['max_oc']) / (df['high'] - df['low'] + 1e-12)
    df['klow'] = (df['min_oc'] - df['low']) / df['open']
    df['klow2'] = (df['min_oc'] - df['low']) / (df['high'] - df['low'] + 1e-12)
    df["ksft"] = (2 * df["close"] - df["high"] - df["low"]) / df["open"]
    df['ksft2'] = (2 * df['close'] - df['high'] - df['low']) / (df['high'] - df['low'] + 1e-12)
    df.drop(columns=['max_oc', 'min_oc'], inplace=True)

    window = [5, 10, 20, 30, 60]
    for w in window:
        df['roc_{}'.format(w)] = df['close'].shift(w) / df['close']

    for w in window:
        df['ma_{}'.format(w)] = df['close'].rolling(w).mean() / df['close']

    for w in window:
        df['std_{}'.format(w)] = df['close'].rolling(w).std() / df['close']

    for w in window:
        df['beta_{}'.format(w)] = (df['close'].shift(w) - df['close']) / (w * df['close'])

    for w in window:
        df['max_{}'.format(w)] = df['close'].rolling(w).max() / df['close']

    for w in window:
        df['min_{}'.format(w)] = df['close'].rolling(w).min() / df['close']

    for w in window:
        df['qtlu_{}'.format(w)] = df['close'].rolling(w).quantile(0.8) / df['close']

    for w in window:
        df['qtld_{}'.format(w)] = df['close'].rolling(w).quantile(0.2) / df['close']

    for w in window:
        df['rank_{}'.format(w)] = df['close'].rolling(w).apply(my_rank) / w

    for w in window:
        df['imax_{}'.format(w)] = df['high'].rolling(w).apply(np.argmax) / w

    for w in window:
        df['imin_{}'.format(w)] = df['low'].rolling(w).apply(np.argmin) / w

    for w in window:
        df['imxd_{}'.format(w)] = (df['high'].rolling(w).apply(np.argmax) - df['low'].rolling(w).apply(np.argmin)) / w

    for w in window:
        shift = df['close'].shift(w)
        min = df["low"].where(df["low"] < shift, shift)
        max = df["high"].where(df["high"] > shift, shift)
        df["rsv_{}".format(w)] = (df["close"] - min) / (max - min + 1e-12)