Determining Optimal Distribution Center locations Using Weighted K-Means

Finding the locations and optimal number of DC’s required in the USA for distribution of the COVID-19 vaccine

Background

Everyone here must have heard of Amazon and its growth in recent years. One of the main reasons for the success of Amazon is its supply chain management. All the people who have ordered at least once from Amazon should be familiar with their popular 1-day shipping. Ever wondered how companies like Amazon can deliver products so quickly to any location in the USA?

The Problem

The total COVID 19 cases count has passed 7 million mark on 24th September in the USA. On the other hand, many countries like the UK are claiming that their COVID -19 vaccine development is in the final stages and will be released by the end of 2020. I thought it will be interesting to find out when the vaccine is finally released how will be distributed across all the hospitals including clinical centres which are treating COVID-19 patients in the USA. If at all the US government is planning to make use of Distribution Center’s to supply vaccination to all hospitals treating COVID-19 patients, Where should the DC’s be located? How many DC’s are needed?

Fig 1. USA hospitals location Data
Determining Optimal Distribution Center locations Using Weighted K-Means 1
Fig 2. Hospital locations in the USA with active COVID-19 case count (Image by author)
  1. Again new centroid points are found by taking the mean of distances

The standard K-means approach would not work because it fails to consider the fact that some regions where hospitals are located have more active COVID-19 cases, which implies having a higher volume demand for the vaccine to be supplied.

How Weighted K-Means differs from Standard K-Means?

Determining Optimal Distribution Center locations Using Weighted K-Means 2
Fig 3. Left Image with no weights vs Right Image with weights (Image by author)
from haversine import haversine
haversine((31.215827,-85.363433),(28.703230,-81.815668))
Determining Optimal Distribution Center locations Using Weighted K-Means 3
Fig 4. Elbow Method to determine the optimal cluster (K=4) (Image by author)
Determining Optimal Distribution Center locations Using Weighted K-Means 4
Fig 5. The Optimal distribution centers locations in the USA (Image by author)
  1. Dallas, Texas
  2. Urbana, Illinois
  3. Petersburg, Virginia.

Conclusion

In this article, an application of the weighted K-means clustering algorithm to determine optimal Distribution locations is demonstrated. To summarize, in this modified K-Means clustering the centroids are calculated by considering the weighted average instead of mean and haversine distance is used instead of euclidean distance.

This article has been pubished from the source link without modifications to the text. Only the headline has been changed.

Source link