Using AI to predict survival probabilities of start-up companies

Research published in the International Journal of Data Science has used machine learning to predict the lifecycle of businesses operating in the digital economy. The work might help firms and policymakers understand enterprise longevity, the rise, the demise or the likelihood of acquisition, in a fast-changing technological landscape.

Shulei Yin of Qilu Normal University in Jinan, Shandong, China, used a gradient-boosting regression tree (GBRT) model to handle the requisite complex, nonlinear relationships within large datasets. She applied it with two tools from survival analysis: the Kaplan-Meier survival curve and the accelerated failure time (AFT) model.

Each tool brings something to the approach. The GBRT model refines prediction accuracy. The Kaplan-Meier curve can estimate the survival probability of firms over time. The AFT model quantifies how external variables, such as competition or enterprise scale, can hasten or slow different phases of a company’s development.

The result of this combination are predictions that offer much greater accuracy than earlier, simpler models. The work is timely given that the traditional business life cycle of start-up, growth, maturity, and decline have become unstable economically speaking. Digital technologies such as cloud computing, artificial intelligence, and big data analytics have lowered entry barriers and condensed innovation timelines.

This means that some companies can scale-up quickly, switch strategies on a whim, or simply lose their market abruptly when the environment and consumer favor shift. Such volatility makes it more difficult for anyone involved in digital activities, whether in the private or public sector actors to predict how things might pan out for a company on which they come to rely for their own operations.

The study thus offers the potential for more certainty in the digital business world. If it is possible to predict which start-ups will thrive and survive, then others can adjust their own strategies accordingly. They can make supply-chain and resource choices based on predicted longevity and avoid signing up for enterprises that might vanish from the scene as abruptly as they appear.