Search

This is a study of a matchmaking system that adaptively adjusts the recommendation model reflecting the user's implicit preference as well as the explicit one. Many matchmaking systems require their users to assign the level of importance, referred to as weight, of a certain attribute such as age, job, and salary when they select dating partners. However, many users do not know the exact level of importance of each attribute and thus, feel burdened to assign weights. Also, even though users explicitly assign weights, they are often in contrast to the users' actual behaviors in many cases. This paper suggests a new matchmaking system called Adaptive Match-Making System (AMMS) that automatically adjusts the weight of each attribute by analyzing the user's previous behaviors. AMMS provides recommendations for newly entered users on the basis of their explicit-weights assigned by users. However, as the user's behavioral records are accumulated, it begins to build the logistic regression model in order to find out the user's implicit weights and reflects them in proportion to the accuracy of the resulting model. The prototype of AMMS is implemented by using Java and the web editor. It is applied to the created artificial dataset based on the real survey results from major matchmaking companies in Korea.

Objective: The paper studies the new learning technique called cost-sensitive case-based reasoning (CSCBR) incorporating unequal misclassification cost into CBR model. Conventional CBR is now considered as a suitable technique for diagnosis, prognosis and prescription in medicine. However it lacks the ability to reflect asymmetric misclassification and often assumes that the cost of a positive diagnosis (an illness) as a negative one (no illness) is the same with that of the opposite situation. Thus, the objective of this research is to overcome the limitation of conventional CBR and encourage applying CBR to many real world medical cases associated with costs of asymmetric misclassification errors. Methods: The main idea involves adjusting the optimal cut-off classification point for classifying the absence or presence of diseases and the cut-off distance point for selecting optimal neighbors within search spaces based on similarity distribution. These steps are dynamically adapted to new target cases using a genetic algorithm. We apply this proposed method to five real medical datasets and compare the results with two other cost-sensitive learning methods-C5.0 and CART. Results: Our finding shows that the total misclassification cost of CSCBR is lower than other cost-sensitive methods in many cases. Even though the genetic algorithm has limitations in terms of unstable results and over-fitting training data, CSCBR results with GA are better overall than those of other methods. Also the paired t-test results indicate that the total misclassification cost of CSCBR is significantly less than C5.0 and CART for several datasets. Conclusion: We have proposed a new CBR method called cost-sensitive case-based reasoning (CSCBR) that can incorporate unequal misclassification costs into CBR and optimize the number of neighbors dynamically using a genetic algorithm. It is meaningful not only for introducing the concept of cost-sensitive learning to CBR, but also for encouraging the use of CBR in the medical area. The result shows that the total misclassification costs of CSCBR do not increase in arithmetic progression as the cost of false absence increases arithmetically, thus it is cost-sensitive. We also show that total misclassification costs of CSCBR are the lowest among all methods in four datasets out of five and the result is statistically significant in many cases. The limitation of our proposed CSCBR is confined to classify binary cases for minimizing misclassification cost because our proposed CSCBR is originally designed to classify binary case. Our future work extends this method for multi-classification which can classify more than two groups.

Case-based reasoning (CBR) methods are applied to various target problems on the supposition that previous cases are sufficiently similar to current target problems, and the results of previous similar cases support the same result consistently. However, these assumptions are not applicable for some target cases. There are some target cases that have no sufficiently similar cases, or if they have, the results of these previous cases are inconsistent. That is, the appropriateness of CBR is different for each target case, even though they are problems in the same domain. Thus, applying CBR to whole datasets in a domain is not reasonable. This paper presents a new hybrid datamining technique called two-step filtering CBR and rule induction (TSFCR), which dynamically selects either CBR or RI for each target case, taking into consideration similarities and consistencies of previous cases. We apply this method to three medical diagnosis datasets and one credit analysis dataset in order to demonstrate that TSFCR outperforms the genuine CBR and RI.

Case-based reasoning (CBR) models often solve problems by retrieving multiple previous cases and integrating those results. However, conventional CBR makes decisions by comparing the integrated result with the cut-off point irrespective of the degree of the adjacency between them. This can cause increasing misclassification error for the target cases adjacent to the cut-off point, since the results of previous cases used to produce those results are relatively inconsistent with each other. In this article, we suggest a new interactive CBR model called grey-zone case-based reasoning (GCBR) that makes decisions focusing additional attention on the cases near the cut-off point by interactive communication with users. GCBR classifies results automatically for the cases placed outside the cut-off point boundary area. On the other hand, it communicates with users to make decision for the cases placed inside the area by verifying characteristics of the dataset. We suggest the architecture of GCBR and implement its prototype.

This paper proposes a regression case based reasoning (RCBR) which applies different weights to independent variables before finding similar cases. The traditional CBR model has focused on finding similar cases from a case base without considering the importance of independent variables. Thus, when extracting similar cases the traditional CBR has put same weights on each independent variable. The proposed regression CBR (RCBR) finds a relative importance of independent variables from the relationship between independent variables and a dependent variable using a regression analysis and puts relative weights using regression coefficients on independent variables. Then, it selects nearest neighbor or similar cases using weighted independent variables through the traditional CBR machine and updates weights dynamically for the next target case and again performs the traditional CBR machine.

This paper proposes a new learning technique which extracts new case vectors using Dynamic Adaptive Ensemble CBR (DAE CBR). The main idea of DAE CBR originates from finding combinations of parameter and updating and applying an optimal CBR model to application or domain area. These concepts are investigated against the backdrop of a practical application involving the prediction of a stock market index.