Machine Learning (ML) algorithms such as classification or regression have been proven to be useful in solving many business challenges. However, some challenges require a deeper understanding of ML and more advanced algorithms to solve. In this post, we will discuss a client project that required a more advanced approach to finding sets of operationally comparable public companies to support the private company valuation process.
Our client had collected several years worth of manually captured sets of operationally comparable public companies, which they wanted to leverage to support their private company valuation process. The challenge was to develop an algorithm that could recommend companies given a starting public company. The client wanted an automated solution to this problem, which would save them time and reduce the risk of errors.
To tackle this challenge, we initially explored potential technical approaches and consulted internally with senior technical advisors. This phase is called an AI Clinic, where we review the technical feasibility and recommend a technical approach. Once we had a recommended technical approach, we developed and iterated on a prototype.
The prototype development phase involved trying hundreds of different configurations in an automated way to find a solution that proved the concept. Ultimately, the technical solution consisted of using a triplet loss model to learn from successive examples of comparable and not-comparable companies.
The triplet loss model is a type of machine learning algorithm used in deep learning. The model learns from successive examples of comparable and not-comparable companies and identifies the underlying features that distinguish them. The algorithm then recommends companies with similar features based on the starting public company.
The outcome of this project was a visual interactive demo that successfully demonstrated a novel capability. The demo showed the client how the algorithm recommended companies given a starting public company. The outcome opened up a much lower risk opportunity to invest in this AI-powered capability.
In conclusion, some business challenges require more advanced techniques, and machine learning algorithms can help solve them. In this case, a triplet loss model was used to recommend companies given a starting public company. This project demonstrated the value of investing in advanced ML techniques and provided a new capability to support the private company valuation process.