This year, Finns do not need to bother going to the market square, as the medal-accustomed hockey nation is predicted by AI to only achieve fifth place. The brightest medals go to North America, with the USA celebrating gold and Canada silver. Adding to the bitterness of the situation for Finland is that Sweden manages to clinch bronze.
AI demonstrated its strength as a forecasting tool when the Finnish Liiga season culminated a couple of weeks ago in exactly the way AI predicted. Read more here.
However, the forecast for the World Championship results has been implemented in a slightly different way. While the Liiga forecast was based on a random forest model, which aids decision-making by calculating the most likely outcome from several different decision trees, the World Championship forecast has been implemented using generative AI.
“It would have been boring to do the same thing again,” says Digia’s CTO Juhana Juppo.
The purpose of the forecasts is to illustrate the possibilities of utilizing AI and to encourage experiments with AI.
“We recommend to our customers that they should not wait to utilize AI, but should get started without delay. Through various experiments, the application targets and utilization methods are found where AI is most beneficial to your own organization,” says Juppo.
Can generative AI be used for forecasting?
Generative AI solutions have become familiar to the general public from services like ChatGPT, which are based on language models and primarily assist in content production or other text processing needs.
Digia’s forecast for the World Championships has been implemented using a RAG-based language model solution. Retrieval-Augmented Generation or RAG is an advanced AI method that combines the generative capabilities of language models and advanced information retrieval techniques to find information: RAG expands the understanding of a mere language model and gives it the ability to search and use documents outside the model to produce more accurate and context-related answers. Read more about how the forecast was implemented here (in Finnish).
The World Championship forecast differs from the Liiga forecast also in terms of the data base. The Liiga forecast was based on the Liiga statistics. Data was available from several tens of different variables, from which the most significant ones were first identified with the help of AI, and then the actual forecast was implemented based on these most significant variables.
There is not as comprehensive data available from the World Championships, and the forecast is purely based on the results of previous games and the success of the countries.
“It is likely that the World Championship forecast will not reach the same accuracy as the Liiga forecast, as the data is not as precise and the method selection is more experimental. On the other hand, when the model has been tested, it has achieved surprisingly good results compared to the random forest model. The forecast is updated every couple of business days during the championships, and we are eagerly watching when AI gets a grip on the correct pattern,” says Juppo.
He reminds that AI is an everyday tool like any other technology. AI solutions can be implemented technically in different ways, and one and the same tool rarely suits all needs.
“It is essential to understand what different implementation methods are suitable for, and what are their strengths and weaknesses. Continuous experimentation is important for learning and finding the tools that suit your own needs. When used correctly, AI is a very good aid for decision-making and many other needs,” says Juppo.
More information:
Juhana Juppo
Chief Technology Officer, Digia Plc.
tel. +358 40 172 2859
juhana.juppo(a)digia.com