Important Ubiquities of 2frame's Work Methodology

The approaches were informed by the literature of the area (if any) from which the project/problem came: conceptualizations, model types, and relevant features (or good proxies) were considered (not exclusively) in the analysis and/or modeling. The idea was to proceed closer to confirmed methodologies (techniques, analyses, modeling/models) arising from the best understanding of the studied phenomena rather than fitting the phenomena into untested methodologies.

In the absence of literature, developed approaches were based on fewer assumptions about the phenomena.

The exchange of information with clients also played an important role in choosing the adopted procedures.

Mathematical/probabilistic techniques were freely used, as long as relevant, in intermediate steps, as approaches per se, and/or to rigorously base the steps, procedures, and techniques designed by 2frame. With this same freedom, optimizations were also "handcrafted" (objective functions and constraints) for the implementation of methods designed for the projects, in pre-processing, in intermediate stages of analysis and modeling, among other uses, guaranteed the pertinence and adequacy of these optimizations with each project as a whole.

Steps common to the initial phase of all projects (non-exclusive):

  • Interactive refinement of customer questions until they become operational.

  • Data gathering/collection and enrichment (buying from external sources), whether necessary.

  • Transformation of the collected data in databases.

  • Analysis of the databases' state and treatment of databases' problems as: poorly filled data, inconsistent data, missing values, among others.

      • Example: Not available data (ownership, secrecy, non-existence), which requires unveiling and finding good proxies.

  • Harmonization of all data in databases.

  • Subsequent types of processing - off-the-shelf or designed (with rigor) by 2frame:

      • Conversions and/or transformations.

      • Groupings and/or clustering.

      • Derivation and/or creation of features and indexes.

      • Standardizations and/or normalizations.

      • Other procedures not previously listed.

This initial phase also encompassed a comprehensive exploratory analysis and visualizations that assisted in the discovery and understanding of the existing relationships and patterns in the information contained in the databases.