The exponential rise in software technologies and its significances has demanded academia-industries to ensure low cost software solution with assured service quality and reliability. A low cost and fault-resilient software design is must, where to achieve low cost design the developers or programmers prefer exploiting source or function reuse. However, excessive reusability makes software vulnerable to get faulty due to increased complexity and aging proneness. Non-deniably assessing reusability of a class of function in software can enable avoiding any unexpected fault or failure. To achieve it developing a robust and efficient reusability estimation or prediction model is of utmost significance. On the other hand, the aftermath consequences of excess reusability caused faults might lead significant losses. Hence assessing cost effectiveness and efficacy of a reusability prediction model is must for software design optimization. In this paper, we have examined different reusability prediction models for their cost effectiveness and prediction efficiency over object-oriented software design. At first to examine the reusability of a class, three key object oriented software metrics (OO-SM); cohesion, coupling and complexity of the software components are used. Furthermore, our proposed cost-efficient reusability prediction model incorporates Min-Max normalization, outlier detection, reusability threshold estimation; T test analysis based feature selection and various classification algorithms. Different classifiers including decision tree (DT), Naive Bayes (NB), artificial neural network (ANN) algorithms, extreme learning machine (ELM), regression algorithms, multivariate adaptive regression spline (MARS) and adaptive genetic algorithm (AGA) based ANN are used for reusability prediction. Additionally, the cost effectiveness of each reusability prediction model is estimated, where the overall results have revealed that AGA based ANN as classifier in conjunction with OO-SM, normalization, T test analysis based feature selection outperforms other state-of-art techniques in terms of both accuracy as well as cost-effectiveness.
