MENU

Want To Know If You Can Be A Fashion Model? There's A Machine-Learning Algorithm For That

Close

When people picture stressful jobs, words like surgeon, firefighter or air traffic controller come to mind. Today, casting directors are in the same league, as an increasing number of aspiring super-models are fighting to be in the spotlight.

What makes the difference between so many gorgeous women? More importantly, how to choose one who truly and fully represent the brand or fashion house? While the general guidelines remain unchanged - models are tall, skinny and beautiful - some tiny details set the best of them apart.

Jaehyuk Park and team at Indiana University in Bloomington set out to find if a machine with a consistent database that runs the proper algorithm can choose and predict which models will become stars. The measure of that, the scientists decided, is the number of runway presences per season.

Fashion Model Directory, an important industry listing website, provided the raw data: age, name, height, hip-to-waist ratio, the affiliation of the model to a top or low level agency. Confirming an industry standard, the facts indicate that these models are sensibly taller and thinner than the average woman of their age.

Park and the team downloaded the data associated with 431 models, all tagged as 'new faces' by the website. On average, each of them only had 3.25 runways during the September 2014 fashion weeks of New York, London, Paris or Milan. The key word here is "on average," since only 24 percent of them stepped on the runway more than once.

The social media presence of the models was another important element of the research. Sixty percent have an Instagram account, which is regarded as being the leading social network in the fashion industry. It is apparent that most of the models working with top agencies are active on Instagram. The collection of metadata associated with the accounts, namely comments and likes, provided useful and somehow surprising results. After computing detailed aspects, such as the general reaction to a post, the team of researchers started to correlate data and facts with the presence on the runways.

Some results were predictable: tall models have better chances, so much that each half-inch of extra height doubles the chances of making it to the runway. Having a contract with a top agency is a major advantage, increasing the chances tenfold. Data supports the importance of social media, even if in unexpected ways: having a lot of Instagram comments raises the chances of runaway presence, but many "likes" will lower it by 10 percent.

The real challenge was to integrate the findings in complex algorithms that would predict the trajectory of models in the fashion weeks of February and March 2015. From a new batch of 15 models, the best algorithm selected six out of eight models that would shine on the runway.

"We find that a strong social media presence may be more important than being under contract with a top agency, or than the aesthetic standards sought after by the industry," they say [pdf].

A few shortcomings of the study exist, however. Firstly, the research team acknowledges that a larger batch of people would be more appropriate. Secondly, there are gender issues, as there is insufficient data on male models. Thirdly, the etalon for success is debatable. Most models are aware that runways differ largely in size and importance: international names such as Chanel and Armani boost the portfolio of a model significantly.

Park and team managed to show that smart usage of technology can contribute to promoting yourself. Their research strengthens the trending field of the "science of success," which unravels how small factors can make or break a career.

Regardless how it seems to the external observer, the fashion industry is a professional, highly competitive field. As such, each advantage, no matter how small, can turn into a game changer. For ascending supermodels, a solid media presence on Instagram might be the golden ticket to the runway.

ⓒ 2018 TECHTIMES.com All rights reserved. Do not reproduce without permission.
Real Time Analytics