Welcome to croston’s documentation!

croston

croston model for intermittent time series

croston.croston.fit_croston(input_endog, forecast_length, croston_variant='original')[source]
Parameters:
  • input_endog – numpy array of intermittent demand time series
  • forecast_length – forecast horizon
  • croston_variant – croston model type
Returns:

dictionary of model parameters, in-sample forecast, and out-of-sample forecast
croston.croston._croston(input_series, input_series_length, croston_variant, w, h, epsilon)[source]
croston.croston._croston_opt(input_series, input_series_length, croston_variant, epsilon, w=None, nop=1)[source]
croston.croston._croston_cost(p0, input_series, input_series_length, croston_variant, epsilon)[source]

example

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
     import numpy as np
     import random
     from croston import croston
     import matplotlib.pyplot as plt


     a = np.zeros(50)
     val = np.array(random.sample(range(100,200), 10))
     idxs = random.sample(range(50), 10)

     ts = np.insert(a, idxs, val)


     fit_pred = croston.fit_croston(ts, 10, 'original') # croston's method

     #fit_pred = croston.fit_croston(ts, 10, 'sba') # Syntetos-Boylan approximation
     #fit_pred = croston.fit_croston(ts, 10, 'sbj') # Shale-Boylan-Johnston


     yhat = np.concatenate([fit_pred['croston_fittedvalues'], fit_pred['croston_forecast']])

     plt.plot(ts)
     plt.plot(yhat)

Indices and tables