Blogwatcher, Inc., which provides advertising and analysis service "Profile Passport" using location information data with Nagoya University, has jointly patented information processing technology that uses profile passport data to model the characteristics of users and areas. Japanese Patent Application No. 2020-209558) has been filed.
In recent years, with the spread of smartphones and wearable terminals equipped with a GPS (Global Positioning System) function, it has become easier to collect location information history.By analyzing data that reflects the daily behavior of each user, it can be used for various purposes such as congestion forecasting, city planning, disaster prevention, and marketing.
Therefore, Nagoya University and Blogwatcher have developed a technology to solve the problems in the conventional technology in order to make it easier to utilize the location information data.
In the conventional technology, when the user's purpose of stay is inferred based on "POI (Point of Interest) = data such as facilities and points including location information" and a stay transition model based on the purpose of stay is created, a determined label is used. Can only be attached.Therefore, there are problems that it is not possible to label POIs that are not registered in the data set, and that it is not possible to determine one label for stores that handle many products.In addition, since the location information history is aggregated into several tens to several hundreds of type labels, the loss of information that originally contained various elements is large.
The patented "vector expression transition model" models the characteristics of users and areas based on the time, day of the week, time zone, and location information while retaining as much as possible the diverse information of the location information history. Technology that can be done.Based on the characteristics of the user's place of stay at each time, "classify areas based on the similarity of user behavior", "analyze characteristics such as lifestyle from user behavior", and "when each user performs machine learning" It is possible to "predict when and where you tend to stay".
In addition, by using this technology, it is not necessary to collect POI information in advance and label it based on it, and it is possible to reduce the labor and cost of area modeling.
