import pandas as pd
import numpy as np
import os.path as op

IMAGE_WIDTH = {5: 15, 20: 60, 60: 180}
IMAGE_HEIGHT = {5: 32, 20: 64, 60: 96}        

Load images by year

year = 2017
images = np.memmap(op.join("./monthly_20d", f"20d_month_has_vb_[20]_ma_{year}_images.dat"), dtype=np.uint8, mode='r').reshape(
                        (-1, IMAGE_HEIGHT[20], IMAGE_WIDTH[20]))

print(images.shape)
images[0]

(67858, 64, 60)

memmap([[  0,   0,   0, ...,   0,   0,   0],
        [  0,   0,   0, ...,   0,   0,   0],
        [  0,   0,   0, ...,   0,   0,   0],
        ...,
        [  0, 255,   0, ...,   0,   0,   0],
        [  0, 255,   0, ...,   0,   0,   0],
        [  0, 255,   0, ...,   0, 255,   0]], dtype=uint8)

Load Labels

• 'Date': The last day of the 20-day rolling window for the chart.
• 'StockID': CRSP PERMNO that identifies the stock.
• 'MarketCap': Market capitalization in dollar, recorded in thousands.
• 'Ret_{t}d': t=5,20,60, next t-day holding period return.
• 'Ret_month': Holding period return for the next month, from the current monthend to the next monthend.
• 'EWMA_vol': Exponentially weighted volatility (square of daily returns) with alpha as 0.05. One day delay is included.

label_df = pd.read_feather(op.join("./monthly_20d", f"20d_month_has_vb_[20]_ma_{year}_labels_w_delay.feather"))
label_df.head()

	Date	StockID	MarketCap	Ret_5d	Ret_20d	Ret_60d	Ret_month	EWMA_vol
0	2017-01-31	10001	133078.0	4.370390e-07	-0.000002	-0.005954	-0.000002	0.000450
1	2017-02-28	10001	133078.0	3.951997e-03	0.002795	0.009953	0.009953	0.000180
2	2017-03-31	10001	133604.0	-7.874612e-03	-0.015749	0.021723	-0.015749	0.000064
3	2017-04-28	10001	131500.0	9.999880e-03	0.016001	0.038072	0.016001	0.000030
4	2017-05-31	10001	133604.0	4.370390e-07	0.021722	NaN	0.023703	0.000015

Plot the sample images

from matplotlib import pyplot as plt
for i in range(5):
    print(i)
    plt.imshow(images[i], cmap='gray')
    plt.show()

0

1

2

3

4