Tableau Conference On Tour 2015 Tokyoで登壇した際の資料です。
サイバーエージェント アドテクスタジオでは複数の広告部門・子会社より構成されており、広告のビッグデータをTableauを利用して可視化し、業務に役立てています。
今回どのような基盤で、どのような解析をしているかについて発表させていただきました。
Tableau Conference On Tour 2015 Tokyoで登壇した際の資料です。
サイバーエージェント アドテクスタジオでは複数の広告部門・子会社より構成されており、広告のビッグデータをTableauを利用して可視化し、業務に役立てています。
今回どのような基盤で、どのような解析をしているかについて発表させていただきました。
This document discusses statistical modeling of behavioral big data. It explains that with observational data, the relationship between cause and effect is difficult to determine from simple comparisons. Statistical modeling involves modeling the data generation process and fitting the model to the data. This allows controlling for other factors that influence both the outcome variable and explanatory variables. The document provides examples of probability distributions commonly used in statistical modeling and discusses issues that arise when the outcome variable is heavy-tailed. It proposes some approaches to dealing with heavy-tailed outcome variables, like using an intervening variable or modeling the increment of the outcome variable.
How Do Newcomers Blend into a Group?: Study on a Social Network GameMasanori Takano
The document describes a study analyzing data from a social network game to understand how newcomers blend into groups. Key findings include:
- Newcomers tended to cooperate with existing group members without receiving prior cooperation, showing they cooperate to construct new relationships.
- Newcomers were more influenced by messaging from others compared to existing members, indicating messaging helps newcomers integrate.
- Both newcomers and existing members were more likely to cooperate with and respond to the cooperation of newcomers over existing members, demonstrating groups accept newcomers.
The analysis provides insights into how humans use reciprocal altruism to construct cooperative relationships with strangers and integrate newcomers into groups.