ScholarBank@NUShttps://scholarbank.nus.edu.sgThe DSpace digital repository system captures, stores, indexes, preserves, and distributes digital research material.Wed, 02 Dec 2020 07:16:07 GMT2020-12-02T07:16:07Z50121- Characterization of medical time series using fuzzy similarity-based fractal dimensionshttps://scholarbank.nus.edu.sg/handle/10635/39154Title: Characterization of medical time series using fuzzy similarity-based fractal dimensions
Authors: Sarkar, M.; Leong, T.-Y.
Abstract: This paper attempts to characterize medical time series using fractal dimensions. Existing fractal dimensions like box, information and correlation dimensions characterize the time series by measuring the rate at which the distribution of the time series changes when the length (or radius) of the box (or hypersphere) is changed. However, the measured dimensions significantly vary when the box (or hypersphere) position is changed slightly. It happens because the data points just outside the box (or hypersphere) are not accounted for, and all the data points inside the box or hypersphere are treated equally. To overcome these problems, the hypersphere is converted to a Gaussian, and thus the hard boundary becomes soft. The Gaussian represents the fuzzy similarity between the neighbors and the point around which the Gaussian is constructed. This concept of similarity is exploited to propose a fuzzy similarity-based fractal dimension. The proposed dimension aims to capture the regularity of the time series in terms of how the fuzzy similarity scales up/down when the resolution of the time series is decreased/increased. Experiments on intensive care unit (ICU) data sets show that the proposed dimension characterizes the time series better than the correlation dimension. © 2003 Elsevier Science B.V. All rights reserved.
Wed, 01 Jan 2003 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/391542003-01-01T00:00:00Z
- Building decision support systems for treating severe head injurieshttps://scholarbank.nus.edu.sg/handle/10635/43182Title: Building decision support systems for treating severe head injuries
Authors: Dora, C.S.; Sarkar, M.; Sundaresh, S.; Harmanec, D.; Yeo, T.T.; Poh, K.L.; Leong, T.Y.
Abstract: In intensive care units, the patients, who are suffering from severe head injuries, usually enter a state of coma. To treat such patients, who are prone to a high risk of mortality, the neurologist adopts certain aggressive and informed decision-making procedures. Designing a decision support system that would automate or enhance this kind of treatment procedure is difficult due to the presence of unclear domain relationships, numerous interacting variables, time-criticality and real-time multiple inputs. We illustrate how the decision analysis framework can be exploited to build a consultative decision support system for the severe head injury management. Specifically, we need (a) to understand the head injury problem with its inherent uncertainties, (b) to structure the problem, and (c) to discern the decision process. The designed system accepts the prognostic factors of a particular patient as the inputs, and subsequently provides the treatment advice as the output. The effectiveness of the treatments is ranked in terms of patient recovery.
Mon, 01 Jan 2001 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/431822001-01-01T00:00:00Z
- Nonparametric techniques to extract fuzzy rules for breast cancer diagnosis problemhttps://scholarbank.nus.edu.sg/handle/10635/78258Title: Nonparametric techniques to extract fuzzy rules for breast cancer diagnosis problem
Authors: Sarkar, M.; Leong, T.-Y.
Abstract: This paper addresses breast cancer diagnosis problem as a pattern classification problem. Specifically, the problem is studied using Wisconsin-Madison breast cancer data set. Fuzzy rules are generated from the input-output relationship so that the diagnosis becomes easier and transparent for both patients and physicians. For each class, at least one training pattern is chosen as the prototype, provided (a) the maximum membership of the training pattern is in the given class, and (b) among all the training patterns, the neighborhood of this training pattern has the least fuzzy-rough uncertainty in the given class. Using the fuzzy-rough uncertainty, a cluster is constructed around each prototype. Finally, these clusters are interpreted as the fuzzy rules that relate the prognostic factors and the diagnosis results. The advantages of the proposed algorithm are, (a) there is no need to know the structure of the training data, (b) the number of fuzzy rules does not increase with the increase of the number of input dimensions, and (c) small number of fuzzy rules is generated. With the three generated fuzzy rules, 96.20% classification efficiency is achieved, which is comparable to other rule generation techniques. © 2001 IMIA. All right reserved.
Mon, 01 Jan 2001 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/782582001-01-01T00:00:00Z
- Fuzzy-rough nearest neighbors algorithmhttps://scholarbank.nus.edu.sg/handle/10635/40756Title: Fuzzy-rough nearest neighbors algorithm
Authors: Sarkar, Manish
Abstract: In this paper the classification efficiency of the conventional K-nearest neighbors algorithm is enhanced by exploiting the fuzzy-rough uncertainty. The simplicity and nonparametric characteristics of the conventional K-nearest neighbors algorithm remain intact in the proposed algorithm. Unlike the conventional one, the proposed algorithm does not need to know the optimal value of K. Moreover, the generated class confidence values, which are interpreted in terms of the fuzzy-rough ownership values, do not necessarily summed up to one. Consequently, the proposed algorithm can distinguish between equal evidence and ignorance, and thus makes the semantics of the class confidence values richer.
Sat, 01 Jan 2000 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/407562000-01-01T00:00:00Z
- Application of K-nearest neighbors algorithm on breast cancer diagnosis problem.https://scholarbank.nus.edu.sg/handle/10635/39076Title: Application of K-nearest neighbors algorithm on breast cancer diagnosis problem.
Authors: Sarkar, M.; Leong, T.Y.
Abstract: This paper addresses the Breast Cancer diagnosis problem as a pattern classification problem. Specifically, this problem is studied using the Wisconsin-Madison Breast Cancer data set. The K-nearest neighbors algorithm is employed as the classifier. Conceptually and implementation-wise, the K-nearest neighbors algorithm is simpler than the other techniques that have been applied to this problem. In addition, the Knearest neighbors algorithm produces the overall classification result 1.17% better than the best result known for this problem.
Sat, 01 Jan 2000 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/390762000-01-01T00:00:00Z
- Fuzzy K-means clustering with missing values.https://scholarbank.nus.edu.sg/handle/10635/39231Title: Fuzzy K-means clustering with missing values.
Authors: Sarkar, M.; Leong, T.Y.
Abstract: Fuzzy K-means clustering algorithm is a popular approach for exploring the structure of a set of patterns, especially when the clusters are overlapping or fuzzy. However, the fuzzy K-means clustering algorithm cannot be applied when the real-life data contain missing values. In many cases, the number of patterns with missing values is so large that if these patterns are removed, then sufficient number of patterns is not available to characterize the data set. This paper proposes a technique to exploit the information provided by the patterns with the missing values so that the clustering results are enhanced. There are various preprocessing methods to substitute the missing values before clustering the data. However, instead of repairing the data set at the beginning, the repairing can be carried out incrementally in each iteration based on the context. In that case, it is more likely that less uncertainty is added while incorporating the repair work. This scheme is further consolidated in this paper by fine-tuning the missing values using the information from other attributes. The applications of the proposed method in medical domain have produced good performance.
Mon, 01 Jan 2001 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/392312001-01-01T00:00:00Z
- Modular pattern classifiers: A brief surveyhttps://scholarbank.nus.edu.sg/handle/10635/39941Title: Modular pattern classifiers: A brief survey
Authors: Sarkar, Manish
Abstract: While solving a complex pattern classification problem, it is often difficult to design a monolithic classifier. One approach is to divide the problem into the smaller ones, and solve each subproblem using a simpler classifier. This kind of divide and conquer policy has motivated the researchers to substitute a modular classifier for the single monolithic classifier. This paper reviews the advantages, issues and various techniques available for designing the modular classifiers.
Sat, 01 Jan 2000 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/399412000-01-01T00:00:00Z
- Measures of ruggedness using fuzzy-rough sets and fractals: Applications in medical time serieshttps://scholarbank.nus.edu.sg/handle/10635/40086Title: Measures of ruggedness using fuzzy-rough sets and fractals: Applications in medical time series
Authors: Sarkar, M.
Abstract: This paper attempts to characterize the medical time series by quantifying the ruggedness of the time series. The presence of two close data points on the time axis implies that these points are similar along the time axis. It creates the fuzzy similarity. Following the principle "similar causes create similar effects", we expect that the magnitudes corresponding to those two data points should also be similar. However, if other features are considered along with the time information, then those two apparently similar data points might look different. Consequently, when the other features are not considered, the magnitudes of those two similar points become different. It makes the time versus magnitude relationship one-to-many. Thus, the closeness creates the fuzziness, the one-to-many relationship creates the roughness, and together they form fuzzy-roughness. If the ruggedness is expressed as the fuzzy-roughness, then in some time series it is observed that the fuzzy-roughness of a part of the time series is similar to that of the whole time series. Specifically, the sealing up of the fuzzy-roughness follows the power law of fractal theory. Experiments on ICU data sets show that the ruggedness measure using the fuzzy-rough set based fractal dimension is more robust than the Hurst exponent, which is used frequently to measure the ruggedness of a fractal time series.
Mon, 01 Jan 2001 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/400862001-01-01T00:00:00Z
- Top-down approaches to abstract medical time series using linear segmentshttps://scholarbank.nus.edu.sg/handle/10635/40085Title: Top-down approaches to abstract medical time series using linear segments
Authors: Sarkar, M.; Leong, T.-Y.
Abstract: This work attempts to abstract medical time series using a minimum number of linear segments such that the integral square error between the abstraction and the data is minimum. The problem is difficult since it involves a multiobjective optimization procedure, and the optimization process is affected by the presence of local minima, noise and outliers. This work proposes a greedy approach, which exploits the local and global information for the optimization. Initially, the number of linear segments needed is estimated roughly by detecting the number of cycles in the data set. Then the tendency of each data point to form bends is measured locally in terms of typicality values. A global consensus in terms of clustering is used to select the breakpoints from all the data points with various typicality values. These breakpoints are utilized to partition the data set. Approximating each partition with a linear segment subsequently forms a crude abstraction. The difference between the original data set and the crude abstraction is exploited as the feedback information such that the crude abstraction can be split further for refinement. The efficacy of the proposed method is demonstrated on some real life intensive care unit (ICU) data sets.
Mon, 01 Jan 2001 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/400852001-01-01T00:00:00Z
- Splice junction classification problems for DNA sequences: Representation issueshttps://scholarbank.nus.edu.sg/handle/10635/41583Title: Splice junction classification problems for DNA sequences: Representation issues
Authors: Sarkar, M.; Leong, T.-Y.
Abstract: Splice junction classification in a Eukaryotic cell is an important problem because the splice junction indicates which part of the DNA sequence carries protein-coding information. The major issue in building a classifier for this classification task is how to represent the DNA sequence on computers since the accuracy of any classification technique critically hinges on the adopted representation. This paper presents the experimental results on seven representation schemes. The first three representations interpret each DNA sequence as a series of symbols. The fourth and fifth representations consider the sequence as a series of real numbers. Moreover, the first, second and fourth representations do not consider the influence of the neighbors on the occurrence of a nucleotide, whereas the third and fifth representations take the influence of the neighbors into considerations. To capture certain regularity in the apparent randomness in the DNA sequence, the sixth representation treats the sequence as a variant of random walk. The seventh representation uses Hurst coefficient, which quantifies the roughness of the DNA sequences. The experimental results suggest that the fourth representation scheme makes sequences from the same class close and the sequences from the different classes far, and thus finds a structure in the input space to provide the best classification results.
Mon, 01 Jan 2001 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/415832001-01-01T00:00:00Z
- Performance of existing prognostic factors for colorectal cancer prognosis - a critical viewhttps://scholarbank.nus.edu.sg/handle/10635/41584Title: Performance of existing prognostic factors for colorectal cancer prognosis - a critical view
Authors: Sarkar, M.; Qi, X.Z.; Leong, P.-K.; Leong, T.-Y.
Abstract: Physicians use several prognostic factors to determine the state of a patient who is in the follow-up program of colorectal cancer. Casting this prognosis problem as a pattern classification problem, we have attempted to find how efficient the prognostic factors are for the classification. We have specifically chosen sixteen most important prognostic factors for conducting experiments. The experimental results show with strong evidence that the chosen prognostic factors have limited discriminatory capability, and hence their use in prognosis may not improve the prognostic efficiency satisfactorily. We present the data analysis and experimental results based on the real data collected from Singapore General Hospital.
Mon, 01 Jan 2001 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/415842001-01-01T00:00:00Z
- Nonparametric techniques to extract fuzzy rules for breast cancer diagnosis problem.https://scholarbank.nus.edu.sg/handle/10635/39628Title: Nonparametric techniques to extract fuzzy rules for breast cancer diagnosis problem.
Authors: Sarkar, M.; Leong, T.Y.
Abstract: This paper addresses breast cancer diagnosis problem as a pattern classification problem. Specifically, the problem is studied using Wisconsin-Madison breast cancer data set. Fuzzy rules are generated from the input-output relationship so that the diagnosis becomes easier and transparent for both patients and physicians. For each class, at least one training pattern is chosen as the prototype, provided (a) the maximum membership of the training pattern is in the given class, and (b) among all the training patterns, the neighborhood of this training pattern has the least fuzzy-rough uncertainty in the given class. Using the fuzzy-rough uncertainty, a cluster is constructed around each prototype. Finally, these clusters are interpreted as the fuzzy rules that relate the prognostic factors and the diagnosis results. The advantages of the proposed algorithm are, (a) there is no need to know the structure of the training data, (b) the number of fuzzy rules does not increase with the increase of the number of input dimensions, and (c) small number of fuzzy rules is generated. With the three generated fuzzy rules, 96.20% classification efficiency is achieved, which is comparable to other rule generation techniques.
Mon, 01 Jan 2001 00:00:00 GMThttps://scholarbank.nus.edu.sg/handle/10635/396282001-01-01T00:00:00Z