Kami 0.4.3

Forecast sales with Entity Embedding LSTM

AM18_SPR20_LondonLAB

Package Description

The package contains one object Kami with four methods in the order of execution:

1. Preprocess()
2. Analyse(n_sample)
3. Vis()
4. Forecast(store_list, product_list, start, end)

To initiate the object Kami, the user is required to supply at least these three arguments:

1. Path to the grouped product sales input data (input_f_path)
2. Path to an intermediary folder to store intermediary data (cache_dir_path)
3. Path to an output folder to store final predictions (output_dir_path)

While Preprocess and Vis methods are executed without any argument, Analyse method can be supplied with an optional argument n_sample which is the number of random samples drawn from the predefined training data.

Forecast method is required to be supplied with four arguments including:

1. A list of stores whose sales are predicted (store_list)
2. A list of products whose sales are predicted (product_list)
3. The start date of the forecast (start)
4. The end date of the forecast (end)

Typical Use Case

---

from Kami import Kami

obj = Kami(input_f_path = 'PATH_TO_SALES_DATA/SALES_DATA.csv',
		output_dir_path = 'OUTPUT_FOLDER/',
		cache_dir_path = 'CACHE_FOLDER/')
obj.Preprocess()
obj.Analyse()
obj.Vis()
obj.Forecast(store_list = ['STORE_A', 'STORE_B'],
		product_list = ['PRODUCT_A', 'PRODUCT_B'],
		start = 'MM/DD/YYYY',
		end = 'MM/DD/YYYY')

---