Case-based reasoning methdos based on statistical analysis

Abstract

Case-Based Reasoning (CBR) has been used in various problem-solving areas such as financial forecasting, credit analysis and medical diagnosis. However, conventional CBR has several limitations that decrease its predictability and availability. In this paper, we address three issues of CBR and suggest new CBR methods for overcoming the limitations by statistical analysis. One of the limitations of conventional CBR is that it uses a fixed number of neighbors without considering an optimal number for each target case, so it does not guarantee optimally similar neighbors for various target cases. This leads to the weakness of lowering the predictability due to deviation from desired similar neighbors. Thus, we suggest a new case extraction technique called Probabilistic Case-Based Reasoning (PCBR) using statistical distribution of distances between cases in this paper. The main idea involves a dynamic adaptation of the optimal number of neighbors by considering the distribution of distances between potential similar neighbors for each target case. In order to do this, our technique finds the optimal distance threshold and selects similar neighbors satisfying the distance threshold criterion. This method overcomes the limitation of conventional CBR, and provides improved classification accuracy. However, CBR methods still have the limitation on not being able to incorporate asymmetric misclassification cost. Thus, our second suggestion is the creation of a new CBR method called Cost-Sensitive Case-Based Reasoning (CSCBR) that can incorporate unequal misclassification cost. PCBR as well as conventional CBR assumes that the cost of type1 error and type2 error are the same, so it cannot be modified according to the error cost of each type. This problem provides major disincentive to apply CBR to many real world cases that have different costs associated with different types of error. CSCBR dynamically adapts both the classification boundary point and the number of...