The scientific community has a special interest in providing solutions to deal with a huge amount of data generated by the Internet, mobile devices, and sensors, among others. One of the foremost research approaches has been in not only SQL (NoSQL) databases, mainly used to handle Big Data. A state of the art review presented in this article let us argue on the need to define modeling techniques to depict how data will be structured in NoSQL databases. The majority of studies have focused on structured data and column oriented databases; thus, we propose an approach for semi-structured data at conceptual and logical modeling levels, using the Unified Modeling Language and key-value databases. The logical model is attained from a class diagram, with the use of transformation rules based on some aspects of the Query View Transformation. Furthermore, our proposal presents a case study concerning data in security log files.