Difference between revisions of "ANLY482 AY2016-17 T2 Group21 : Finals"
Line 60: | Line 60: | ||
==Survival Analysis on Stock Out Time== | ==Survival Analysis on Stock Out Time== | ||
− | [[File:Tableabc.png]] | + | [[File:Tableabc.png|300px]] |
To account for non-stockout products, we perform our survival analysis of product stock-out-time with the following definition: | To account for non-stockout products, we perform our survival analysis of product stock-out-time with the following definition: |
Revision as of 17:37, 23 April 2017
Exploratory | Mid-Term | Finals |
---|
Survival Analysis
Due to the censored demand identified during our exploratory analysis, using survival analysis provides a way for us to handle such hidden values. Survival analysis will be performed using the JMP built-in survival functions. We will be using two features:
1. Basic survival function
- Applies Kaplan-Meier estimator to account for censored values
2. (Cox) proportional hazards fit
- Fits a linear model between predictors (explanatory variables) and the hazard function.
- Parameters estimates show how predictors affect the hazard function.
Survival Analysis on Stock Out Time
To account for non-stockout products, we perform our survival analysis of product stock-out-time with the following definition:
- Subject: A product identified by name and size
- Time to event: Time in days for a product to stockout
- Censor: 0 if product stockout, 1 otherwise
Figure 8 Here
The above analysis shows a more accurate median time of stockout of 18 days, which is longer than our sponsor’s target of achieving a stockout period of 7 days. From the survival plot, we can also see that 68% of all products still remains on the shelf after 7 days after launch. It is, therefore, useful to understand which products groups have a longer stockout periods. We further add groupings by category to our analysis.