C. Carter and J. Catlett. Assessing credit card applications using machine learning. IEEE Expert, Fall:71--79, 1987.  Home/Search   Document Not in Database   Summary   Related Articles   Check

.... where the kNN method outperforms them (as a confirmation of this, we notice that kNN actually outperforms tree bagging significantly on the WAVEFORM dataset) Another recent class of proposals more related to our local approach and similar in spirit to the early work of Carter and Catlett [14], consists in using continuous transition regions instead of crisp thresholds. This leads to overlapping subsets at the successor nodes and weighted propagation mechanisms. For example, in a fuzzy decision tree, fuzzy logic is used in order to build hierarchies of fuzzy subsets. Wehenkel ( 9] ....

C. Carter and J. Catlett. Assessing credit card applications using machine learning. IEEE Expert, Fall:71--79, 1987.

....rule, and reported that the resulting trees are more accurate than univariate trees. Soft splits Soft splits are those in which the split function is smoothed to allow for a continuous transition across a threshold . These have been considered by a number of authors. Carter and Catlett [31] defined upper and lower cut points adjacent to the threshold T and used linear interpolation to smooth 2.4. REVIEW OF RELATED LITERATURE 25 between the following points: # , # T , choose T10 . 2.37) The choice of was made using methods derived via ....

Carter, C. and Catlett, J. [1987], `Assessing credit card applications using machine learning', IEEE Expert 2(3), 71--79.

....to use the built tree [Maher et al..93] However, this kind of methods produces large trees and we can consider that it is difficult to rely on induction from a single numerical datum. Other methods introduce a partitioning step where we search for cut points on a numerical universe, either before [Carter et al..87, Catlett91] or during the construction of the tree [Breiman et al..84] But for values near the cut points, the generated tests of the tree are often very imprecise in nature. Therefore, one possible way of improvement lies in the use of fuzzy concepts when classifying a new case, in order to ....

Carter (C.) et Catlett (J.). -- Assessing credit card applications using machine learning. IEEE Expert, vol. Fall Issues, 1987.

....represent 690 MasterCard applicants of which 307 are declared as positive and 383 as negative. The data contain 37 records with missing entries. Each record consists of 15 attributes, of which 4 are Boolean, 5 nominal 21 (that is, descriptive) and 6 rational. For prior computational results, see Carter and Catlett (1987), and Boros et al. 1996) The representation of the 15 attributes requires a total of 67 logic variables. With this transformation, A and B had one record in common. We have removed such record from A to achieve consistency. Table 2 summarizes the results. Table 2: Australian Credit Card 689 ....

Carter, C., Catlett, J. 1987. Assessing Credit Card Applications Using Machine Learning. IEEE Expert Fall 1987 pp. 71-79.

....RECYCLING DECISION TREES IN NUMERIC DOMAINS Informatica 24 page 194 205 5 Table 1: Gauss A. The target contexts have shifted means. source target 1 target 2 target 3 target 4 negative examples ( 0,0] 1) 1,0] 1) 2,0] 1) 4,0] 1) 10,5] 1) positive examples ( 2,0] 2) 3,0] 2) 4,0] 2) [6,0],2) 12,5] 2) C4.5: re used 78.0 75.0 60.8 48.0 47.5 C4.5: re learned 78.0 78.0 78.0 78.0 78.0 DT 2T: re used 81.0 75.3 61.3 47.5 52.5 DT 2T: re learned 81.0 81.0 81.0 81.0 81.0 DT 2T: post tuned 81.0 80.8 80.5 80.8 81.0 Table 2: Gauss B. The target contexts have altered standard ....

Carter, C. and Catlett, J. (1987). Assessing Credit Card Applications Using Machine Learning, IEEE Expert, Fall issue, 71-79

....0, 1 12 class is stemmed phrase a keyphrase, based on match with stemmed form of human generated keyphrases 0, 1 Turney 16 trol the number of feature vectors that are classified as belonging in class 1. Therefore we ran C4.5 with the p option, which generates soft threshold decision trees (Carter and Catlett, 1987; Quinlan, 1987, 1990, 1993) Soft threshold decision trees can generate a probability estimate for the class of each vector. For a given document, if the user specifies that K keyphrases are desired, then we select the K vectors that have the highest estimated probability of being in class 1. In ....

....or any of the keyphrases. Learning Algorithms for Keyphrase Extraction 39 10. Discussion We have presented two approaches to the task of learning to extract keyphrases from text. The first approach was to apply the C4.5 decision tree induction algorithm (Quinlan, 1993) using soft thresholds (Carter and Catlett, 1987; Quinlan, 1987, 1990, 1993) bagging (Breiman, 1996a, 1996b; Quinlan, 1996) and stratified sampling (Deming, 1978; Buntine, 1989; Catlett, 1991; Kubat et al. 1998) The experiments support the claim that bagging is helpful for this task, but stratified sampling is not helpful. Our experience ....

Carter, C., and Catlett, J. (1987). Assessing credit card applications using machine learning. IEEE Expert, Fall issue, 71-79.

....restrictive property) the problem cannot in general be decomposed this way, and the observed values of f(x) for unobserved x are in general restricted. In [2] we consider some other possible restrictions on f (when it is a boolean function) More examples for similar problems can be found in [3, 14, 15]. In this paper we show that a network flow model can be used to determine the best approximation f when the partial ordering of the outcomes is an interval order. An interval order is one in which when every element can be associated with an interval of real numbers, such that ff fi if and only ....

Carter, C. and J. Catlett, "Assessing Credit Card Applications Using Machine Learning", IEEE Expert (Fall 1987), 71-79.

....consists of 690 records of MasterCard applicants, 307 of which are classified as positive and 383 as negative. While 37 records have some missing data, they were not omitted from our analysis. Each record consists of 15 attributes (4 Boolean, 5 nominal, and 6 numerical) C. Carter and J. Catlett [4] reported experiments on this dataset using a decision tree technique, and obtained a correct prediction rate of 85.5 , using a training set of 71 . In a testing set of 200 observations, 171 points were classified correctly, 18 points were misclassified, and 11 points were not classified at all. ....

....real data taken from various fields of applications. The evaluation of the performance of 18 IDIAP RR 96 05 LAD classifier Other approaches 50 training 80 training best found used for DataBase AVG STD AVG STD AVG STD training reference Australian Credit Card 85.4 1.2 85.5 2. 6 85.5 71 [4] Boston Housing 84.0 1.6 85.2 3.0 83.2 3.1 80 [16] Breast Cancer (Wisconsin) 96.9 0.9 97.2 1.3 96.2 0.3 80 [16] Congressional Voting 96.2 1.1 96.6 1.8 95.6 66.6 [12] Diabetes 71.9 1.9 72.3 2.4 76 75 [19] Heart Disease (Cleveland) 82.3 1.7 83.8 5.2 80.6 3.1 [18] not available Table ....

Chris Carter and Jason Catlett. Assessing credit card applications using machine learning. IEEE Expert, pages 71--79, Fall 1987.

....: 14 1 Introduction A common inference task consists of making a discrete prediction about some object given other details about the object. For instance, in financial credit assessment as discussed by Carter and Catlett (Carter and Catlett, 1987) we wish to decide whether to accept or reject a customer s application for a loan given particular personal information. This prediction problem is the basic task of many expert systems, and is referred to in AI as the classification problem (where the prediction is referred to as the ....

Carter, C. and Catlett, J. (1987). Assessing credit card applications using machine learning. IEEE Expert, 2(3):71--79.

....rule, and reported that the resulting trees are more accurate than univariate trees. Soft splits Soft splits are those in which the split function is smoothed to allow for a continuous transition across a threshold t j . These have been considered by a number of authors. Carter and Catlett [31] defined upper and lower cut points t j and t Gamma j adjacent to the threshold t j and used linear interpolation to smooth 2.4. REVIEW OF RELATED LITERATURE 25 between the following points: if x j t Gamma j , choose R l with probability 1; if x j = t j , choose R l or R r with ....

Carter, C. and Catlett, J. [1987], `Assessing credit card applications using machine learning', IEEE Expert 2(3), 71--79.

....consists of 690 records of MasterCard applicants, 307 of which are classified as positive and 383 as negative. While 37 records have some missing data, they were not omitted from our analysis. Each record consists of 15 attributes (4 Boolean, 5 nominal, and 6 numerical) C. Carter and J. Catlett [4] reported experiments on this dataset using a decision tree technique, and obtained a correct prediction rate of 85.5 , using a training set of 71 . In a testing set of 200 observations, 171 points were classified correctly, 18 points were misclassified, and 11 points were not classified at all. ....

....LAD, those observations that were not classified by the LAD classifier were counted as errors. Page 22 RRR 22 96 LAD classifier Other approaches 50 training 80 training best found used for DataBase AVG STD AVG STD AVG STD training reference Australian Credit Card 85.4 1.2 85.5 2. 6 85.5 71 [4] Boston Housing 84.0 2.1 85.2 3.4 83.2 3.1 80 [16] Breast Cancer (Wisconsin) 96.9 0.9 97.2 1.3 96.2 0.3 80 [16] Congressional Voting 96.2 1.4 96.6 2.1 95.6 66.6 [12] Diabetes 71.9 2.8 72.2 4.3 76 75 [19] Heart Disease (Cleveland) 82.3 1.7 83.8 6.0 80.6 3.1 [18] Table 3: Correct Prediction ....

Chris Carter and Jason Catlett. Assessing credit card applications using machine learning. IEEE Expert, pages 71--79, Fall 1987.

.... but cannot be represented in the form of a decision tree; they were induced by the PRISM program that infers rules directly from the data without going through a decision tree stage first [Cendrowska, 1987] A more realistic example is the use of ID3 for assessing credit card applications [Carter Catlett, 1987]. Attributes such as bank balance, monthly expenditure, monthly disposable income, employment status, home status, time at address, age of car, and so on are used to form a decision tree to determine whether an applicant is creditworthy or not. Sample cases can be obtained from past applications ....

....data, and so the decision tree is used to generalize from the training set rather than simply to summarize it. Also, creditworthy noncreditworthy categorization is far more subjective than the disease classification in the soybean example, and some noise and inconsistency is only to be expected. Carter Catlett [1987] conclude that machine learning techniques compete well with existing methods for credit assessment by specially trained experts. The ID3 procedure has been thoroughly studied and extended in several different directions. Machine learning by the induction of decision trees has been given a sound ....

Carter, C. and Catlett, J., 1987: "Assessing credit card applications using machine learning." IEEE Expert 2 (3), pp. 71--79; Fall.

....can work with a representative sample from the training set, called windowing, which however, cannot guarantee to yield the same decision tree as would be obtained from the complete set of training 3 examples. In this case the decision tree would be unable to classify all examples correctly [Carter and Catlett, 1987]. AQ is another well known inductive learning algorithm. The original AQ does not handle uncertainty very well. Existing implementations, such as AQ11[Michalski and Larson, 1978] AQ15[Michalski et al. 1986] handle noise with pre and postprocessing techniques. The basic AQ algorithm however, ....

Carter, C., and Catlett, J. (1987). "Assessing Credit Card Applications Using Machine Learning", IEEE Expert, 2(3), 71-79.

....hand written symbols on a pen pad computer. Predicting suitable habitat for Tsetse flies [29] Financial trading [28] Species and sex of Leptograpsus crabs [7] Forensic studies of DNA, fingerprints [8] glass fragments [30, 32] Identifying incoming missiles. Credit allocation rules [10]. Damage to clothes by washing powders [9] In all these of tasks there is a predefined set of classes of patterns which might be presented, and the task is to classify a future pattern as one of these classes. Such tasks are called classification or supervised pattern recognition 1 . Clearly 1 ....

Carter, C. and Catlett, J. (1987) Assessing credit card applications using machine learning. IEEE Expert 2(3), 71--79.

No context found.

C. CARTER AND JASON CATLETT. Assessing credit card applications using machine learning. IEEE Expert, Falh71-79, 1987.

No context found.

C. Carter and Jason Catlett. Assessing credit card applications using machine learning. IEEE Expert, Fall:71--79, 1987.

No context found.

Carter, C. and Catlett, J. (1987). Assessing Credit Card Applications Using Machine Learning, IEEE Expert, Fall issue, 71--79

No context found.

Carter, C., and Catlett, J. (1987). "Assessing Credit Card Applications Using Machine Learning", IEEE Expert, 2(3), pp.71-79.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC