Recent findings in the field of neuroimaging have greatly enriched our understanding of the process of human inductive reasoning. However, as an extremely complex cognitive activity, inductive reasoning involves multiple cognitive functions such as attention, memory, executive function, symbol processing, and fluid intelligence, and requires the collaborative participation of multiple brain regions. Therefore, achieving a unified and transparent understanding of human inductive reasoning from different datasets presents formidable challenges. This paper proposes a multi-loop based integration analysis framework of multiple source brain data. By employing representational similarity analysis and multi-task functional neuroimaging analyses, this framework combines internal evidence, i.e., multi-source functional neuroimaging data, and external evidence, i.e., knowledge from published neuroimaging articles, to reveal different neural mechanisms of inductive reasoning in DLPFC and SPL. Experimental results reveal the intricate uncertainty surrounding brain localization in human reasoning.