Evaluation of estimation models using the Minimum Interval of Equivalence.

Graphical abstractDisplay Omitted HighlightsDefinition of a new measure for evaluating estimation models.The measure is based on the concept of Equivalence Hypothesis Testing.Application of the measure to estimations by different soft computing methods.Construction of probability intervals for each estimation method.Genetic programming and linear regression provide the best intervals. This article proposes a new measure to compare soft computing methods for software estimation. This new measure is based on the concepts of Equivalence Hypothesis Testing (EHT). Using the ideas of EHT, a dimensionless measure is defined using the Minimum Interval of Equivalence and a random estimation. The dimensionless nature of the metric allows us to compare methods independently of the data samples used.The motivation of the current proposal comes from the biases that other criteria show when applied to the comparison of software estimation methods. In this work, the level of error for comparing the equivalence of methods is set using EHT. Several soft computing methods are compared, including genetic programming, neural networks, regression and model trees, linear regression (ordinary and least mean squares) and instance-based methods. The experimental work has been performed on several publicly available datasets.Given a dataset and an estimation method we compute the upper point of Minimum Interval of Equivalence, MIEu, on the confidence intervals of the errors. Afterwards, the new measure, MIEratio, is calculated as the relative distance of the MIEu to the random estimation.Finally, the data distributions of the MIEratios are analysed by means of probability intervals, showing the viability of this approach. In this experimental work, it can be observed that there is an advantage for the genetic programming and linear regression methods by comparing the values of the intervals.