Demand clustering

Clustering distributes a set of objects into various groups so that objects in the same group are more similar to each other than to objects in other groups.

Supply chain designers study future product demand to help determine whether their capacity and network layout will meet upcoming needs. This analysis is often impractical at the individual SKU level, where the relationships between products is not always obvious. The method of filtering time series by demand classification is simply not enough. Demand clustering allows you to group similar products based on certain demand attributes and also gain a deeper understanding of the similarities between products.

(Source:Clustering and Dimensionality Reduction: Understanding the “Magic” Behind Machine Learning)

Clustering examples include:

• Market Research - partition customers into market segments for product positioning and new product development
• Recommender Systems - predict whether a consumer will like a certain product based on the preferences of other consumers in the same cluster
• Social Network Analysis - detect communities in a large network of people
• Anomaly Detection - identify data outliers by examining smaller clusters that are away from most other clusters

Refer to the following for additional information on clustering:

• Time series clustering
• Feature extractions for clustering
• Scaling and outlier adjustments of clustering features
• Choosing a clustering algorithm
• Managing cluster definitions