INNER CODE UNIT · Python
single_symbol_image
RichardS0268/CNN-for-Trading · dataset.py:17
def single_symbol_image(tabular_df, image_size, start_date, sample_rate, indicators, show_volume, mode):
''' generate Candlelist images
parameters: [
tabular_df -> pandas.DataFrame: tabular data,
image_size -> tuple: (H, W), size shouble (32, 15), (64, 60)
start_date -> int: truncate extra rows after generating images,
indicators -> dict: technical indicators added on the image, e.g. {"MA": [20]},
show_volume -> boolean: show volume bars or not
mode -> 'train': for train & validation; 'test': for test; 'inference': for inference
]
Note: A single day's data occupies 3 pixel (width). First rows's dates should be prior to the start date in order to make sure there are enough data to generate image for the start date.
return -> list: each item of the list is [np.array(image_size), binary, binary, binary]. The last two binary (0./1.) are the label of ret5, ret20
'''