Credit Invisible is one key area that many countries put much effort to solve in decades. According to the 2020 World Bank statistics, for example, there are over 500 million Chinese and 45 million American, classified as credit invisible who don’t have banking and finance history in bank or credit bureau, making them difficult to borrow money from financial institution. Previous studies adopted different non-financial information to evaluate one’s credit worthiness and status to address this issue. However, they provide little information about how real mobile user interactions can be used to solve this issue in inclusive finance. This paper proposes a novel data generative framework to fusion APP data, call detail record data and SMS data with a total of 4,689 attributes derived from a large-scale mobile dataset. We then construct a unique set of mobile behavior-driven credit risk factors based on statistical diversity, intensity, consistency, and regularity of mobile user behavior characterizing user preferences, attitudes, geolocation, and temporal patterns. Empirical analysis demonstrates that the newly discovered mobile behavior factors are useful as new inputs for credit scoring and proves the factors representing new source of positive and negative credit information. Decision tree analysis and Quantile regression are conducted to validate effect of these factors to credit default. It facilitates credit assessment based on non-financial data for the credit invisible people, which promoting inclusive finance to larger community in society. We also analyze implications of mobile user characterization findings in relation to credit default which helps decision makers to optimize credit policy and product design.