Authors
Leo Antunes, Alexandre Correa, & Marcio Barros
Abstract
The flexibility and ease-of-use of spreadsheets allow users with little CS background to build computational solutions to solve their problems. However, inexperienced users may build erroneous spreadsheets. Modeling is commonly used in software development to allow users to visualize, communicate, and validate different aspects of a system before starting its construction.
This work proposes a Model-Driven Engineering (MDE) approach that automatically generates spreadsheets from conceptual models aiming to reduce the number of errors introduced by users while using spreadsheets. Initially, we designed and executed an empirical study to identify the most common errors users commit when developing spreadsheets.
The results of this study have shown that even users who classify themselves as experienced spreadsheet developers commit a large number of errors while developing a spreadsheet. Also, it has shown that formula-related errors are the most common errors introduced by users (55%).
A second study involved humans using a spreadsheet generated by our approach and showed initial evidence that the approach can build spreadsheets that avoid the introduction of several types of errors often present in spreadsheets manually created by users. In this sense, we observed 98% correct answers to 234 questions involving spreadsheet manipulation.
Sample
We asked a group of 20 subjects with differing experience in spreadsheet development to build a spreadsheet for an inventory control system.
This table reports the number of errors found on spreadsheets developed by our subjects, according to the selected taxonomy and phases of the development process.
Most errors found in the spreadsheets developed by our subjects were observed in the formula development stage (55%).
Publication
2015, 29th Brazilian Symposium on Software Engineering, September, pages 140-149
