Hybrid approaches for classification under information acquisition cost constraint

We address a problem of classification with information acquisition cost constraint (CIACC). The objective of the CIACC problem is to develop a classification function that maximizes correct classifications under the user defined information acquisition cost constraint. We propose hybrid simulated annealing and neural network (SA-ANN), and tabu search and neural network (TS-ANN) procedures to solve the CIACC problem. Using a simulated and a real-world data set from medical domain, we show that the proposed hybrid procedures solve the CIACC problem. The results of our experiments indicate that the performance of hybrid approaches is sensitive to the data distribution, and memory-based hybrid tabu search approaches may perform as well as or better than probabilistic hybrid simulated annealing approach.