Is there a “Best” data modeling solution?

by Caezar M

in this article the authors try and discover the best method for modeling data from the top three most widely used models. the models in this study are Relational, Extended Entity Relation, and Object Oriented modeling structures. the main goal of this study was to determine which model provided the best abstraction level and the model complexity. the authors also set out to test the validity of the “Abstraction Level Hypothesis” which states that at higher abstraction levels better queries result as opposed to those written at lower abstraction levels. their main goal being why the three constructs offer different modeling performance for the same problem. the authors definition of performance being the level of complexity in relation to the number of errors produced in the models. the authors had three hypotheses which amount to the following, EER and OO modeling will yeild better better modeling correctness and performance than the relational model. in their study of 75 freshman computer students they found that models written at higher abstraction levels supported their hypothesis in all categories except two where correctness in the relational model was higher than the OO model. these categories were binary m:m relationships and ternary relationships. they also found that EER outperform Relational models in 4/7 categories and the remaining 3 are equal. with these findings they managed to prove that in terms of correctness and complexity that the OO model is superrior to both the EER and Relational models however not by a large margin. in comparason with other major studies in the same area they find that their studies are consistent with other studies and the variance is not statistically significant. as for the “Abstraction Level Hypothesis” they proved that this held true in most situations across all the models.


i thought this was a great article because it shows that yes, some modeling methods are better than others in understanding how data is related, understood, and correctly executed, buthow that no method is vastly superrior to the other. every method will have its strengths and its weeknesses but overall there is no one super method that is the answer to every problem. i believe that the authors did a great job in showing what they believed should be the case and how they offered a solution that put each modeling method on a relatively even platform to show how one compared to the other with little bias as to how each method worked. knowing how OO modeling worksi do appreciate its simplicity and effectiveness when it comes to programming and i also have a respect for relational modeling because it offers a better clarity as to just how everything is interrelated. so after reading this study its good to know that these methods while offering their best are all better at something than the other but all help us better understand what we are doing exactly.

Hock-Hai Teo, Hock Chuan Chan, Kwok Kee Wei (2006). Performance Effects of Formal Modeling Language Differences: A Combined Abstraction Level and Construct Complexity Analysis. IEEE Transactions on Professional Communication,  49(2), 160-171

1 thought on “Is there a “Best” data modeling solution?”

  1. Really interesting research article! I also like the way you summarized the article – starting from the background, the research problem and hypothesis, the research method, result and discussion, every aspect in one or two sentences in your own words. It is also interesting experiment design – they controlled the level of complexity of the problems that need to be modeled, and measured the correctness of students’ modeling results, although i think the evaluation of performance is a little bit overlapped because it is measured by complexity relative to the number of errors. I would like you to present this to the class next Monday. Thanks.

Comments are closed.