Signed graphs for portfolio optimization

Abstract

This project aimed to study and analyze the representation of signed networks based on degree of balance. It provides a way of looking at a product portfolio in the ecommerce & retail context and shows results of product portfolio representation & optimization from the project’s idea. To achieve this, the structure of the customer product relationships were mapped into two types of signed graph triangles: two customer one product, and two product one customer; where the customers and products are nodes; and the edges or links between the nodes can be customer-customer, customer-product, or product- product. Next, the notion of balance and unbalanced triangles were formed. On this basis, the project went on to explore the notion with real world data from Amazon and produced a surprising fit: a high degree of balance in the product portfolio of smartphones. An approach to optimize the portfolio using sensitivity analysis was then introduced, outlining the overall portfolio health based on the degree of balance indicator. Lastly, interactive network visualization of the portfolio was also added to showcase the possibilities of analyzing such a product portfolio: feature rich and intuitive. In conclusion, the project’s notion provides a novel way of representing product portfolio, sets a different basis from analyzing such product portfolios, and introduces opportunities to further analyze and optimize the portfolio in a holistic and intuitive way.