The intensifying need to incorporate knowledge extracted from qualitative information into banks' lending decision has been recognized in recent times, particularly for micro lenders. In this study, the multi-faceted credit information is captured in an integrated form using XML to facilitate the discovery of knowledge models encompassing a broad range of credit risk related aspects. The quantitative and qualitative credit data obtained from the industry partner describes existing lender profiles. The experiments are performed to discover classification models for the performing or non-performing lenders in one problem setting, and the duration of payment delay in another. The results are compared with a common credit risk prediction setting where qualitative data is excluded. The findings confirm the role of domain experts' knowledge as well as qualitative information on loan performance assessment, and describe a number of rules indicating refinement of the banks' lending policy requirement.