From 12b8be2f6a2b08a38c048ac3615ecda0882d6036 Mon Sep 17 00:00:00 2001 From: juanitorduz Date: Thu, 19 Sep 2024 00:29:50 +0200 Subject: [PATCH] minor improvements --- croston_numpyro/index.html | 2 +- html/timeseries_numpyro.html | 11 +++++++---- 2 files changed, 8 insertions(+), 5 deletions(-) diff --git a/croston_numpyro/index.html b/croston_numpyro/index.html index 71c8c92..1e7204e 100644 --- a/croston_numpyro/index.html +++ b/croston_numpyro/index.html @@ -177,7 +177,7 @@

Idea of Croston’s Method

As mentioned earlier, Croston’s method is based on the idea of separating the original series \(y_{t}\) into the demand size and the demand interval.

Lets see how we can do this:

# Demand time series
diff --git a/html/timeseries_numpyro.html b/html/timeseries_numpyro.html
index 929505b..513a93a 100644
--- a/html/timeseries_numpyro.html
+++ b/html/timeseries_numpyro.html
@@ -8,7 +8,7 @@
 
 
 
-  
+  
 
   
   Time Series Forecasting with NumPyro
@@ -669,14 +669,17 @@ 
Croston’s Method

 
 
 

-
 The method is based on the idea of separating the demand size and the demand interval, and then forecasting them separately using simple exponential smoothing. 

+
 The method is based on the idea of separating the demand size \(z_t\) and the demand interval \(p_t\), and then forecasting them separately using simple exponential smoothing. 

 

 
 
 
    
-
  • To generate the demand size time series \(z_t\) we simply keep the non-zero values of the original series.
    • 
-
  • To generate the demand interval time series \(y_t\) we keep the time between non-zero values of the original series.
    • 
+
  • \(z_t\): keep the non-zero values of \(y_t\).
    • 
+
  • \(p_t\): keep the time between non-zero values of \(y_t\).
    • 
 

+
\[
+\hat{y}_{t+h} = \frac{\hat{z}_{t+h}}{\hat{p}_{t+h}}
+\]

 

 
Croston’s Method for Intermittent Time Series Forecasting in NumPyro