NTT DOCOMO, INC. announced today a new artificial intelligence (AI) technology called the Dual-view Adaptive Retrieval-augmented Tweedie model. This DOCOMO artificial intelligence technology can make high-accuracy predictions even with a small amount of historical data. And it solves the “cold-start problem.” This problem happens when AI does not have enough historical data to make reliable recommendations or predictions. One is the launch of a new service. DOCOMO says the model will facilitate new businesses across a range of industries, including digital out-of-home (DOOH) advertising.
The model was also accepted for publication in the Association for Computing Machinery (ACM). This paper is accepted for presentation at the 20th ACM Conference on Recommender Systems (ACM RecSys 2026). The conference is a premier global event. Thus the acceptance is an acknowledgment of the novelty and performance of the model.
The Cold-Start Problem and Its Importance
When providers introduce a new service, they often lack historical data. If they expand an existing service into a new field or area, they soon run into the same difficulty. This makes predictions based on AI unreliable. This gap undermines AI-powered recommendation features and advertising-performance forecasts. This leaves providers struggling to recommend or tailor advertising to users’ interests. That makes it harder to keep users, harder to build new businesses
Two Important Features of the Model
To this end, the model incorporates two major features.
First, the model employs the “Tweedie distribution”. This statistical-probability distribution is flexible to capture complex real world data, like data with many zeros or large variation. Specifically, it’s about the rules used to train AI models. While the traditional training rules are based upon the “Gaussian distribution”. It assumes the data points group around the mean and are symmetrically distributed in a bell curve. Values close to the mean are more common than values that are farther away. But conventional artificial intelligence may struggle to learn well in complex environments. Some data, like social media engagement, can be very different in peak vs. off-peak times. Data can also be very different. The model instead uses the Tweedie distribution to allow for correct training of models. It can predict even if the data is unevenly distributed.
Second, the model employs “nearest neighbors.” Such cases of DOCOMO artificial intelligence are useful when there is limited historical data available for the prediction target. The model automatically detects and learns common features in this data such as location, time and other features. such as other stores in the same area other products in similar categories etc. Then the model integrates this information into its predictions.
Advantages for Service Providers and Users
The model allows service providers to deliver accurate AI based predictions and recommendations from launch. They no longer have to wait for large quantities of new data. So this DOCOMO artificial intelligence technology could significantly impact to accelerate new business launches.
For example, in DOOH advertising the model can predict the number of impressions generated by newly installed digital signage on the first day. This is true even in locations with large variations in the number of people walking by, such as major city train stations. The predictions are based on limited historical data, and providers can price advertising slots and begin selling them soon after installation.
Additionally, the model makes it easy for end users to receive content and information that matches their interests from the moment they begin using a service. All of this enhances the service experience overall.
DOCOMO artificial intelligence technology will be used in field trials with DOOH businesses in Japan and overseas to assess the effectiveness of the model by March 2027. Eventually, the company hopes to enable commercial deployment around the world.
Explore IT Tech News for the latest advancements in Information Technology & insightful updates from industry experts!
News Source: Businesswire.com
