Chun-Lang Chang

Applying decision tree and neural network to increase quality of dermatologic diagnosis

Skin diseases are common to children and adults. Many factors influence the onsets of these diseases, and each age group usually has its different symptoms. In the humid, damp, and hot weather conditions of Taiwan, bacteria and molds grow best and fast. Also, exposures to excess amounts of ultraviolet radiations in the sunlight will make skin sensitive, easy to be infected, and possibly cause skin problems. In addition to the external infections, internal sebaceous glands, dead skin, sweats, mixed with dusts and other unwanted secretions can cause other serious skin diseases. Although skin diseases are easier to detect, and diagnosing symptoms and deciding treatment plans are not as complex as other internal diseases, many people often ignore the importance of them. In fact, even a small spot on the skin might cause skin cancer. This study conducted five experiments focusing on six major skin diseases as its research subjects. It uses decision tree of data mining combining with neural network classification methods to construct the best predictive model in dermatology. The results show that using neural network model has the highest, 92.62%, accuracy in prediction. Using sensitivity analysis combining with decision tree model, on the contrary, has the least accuracy, which is 80.33%. Based on this result, the AI classification technology can serve as important and useful references in diagnosis for physicians to avoid unnecessary medical waste and enhance health care quality.

A STUDY ON FLOWSHOP SCHEDULING PROBLEM COMBINING TAGUCHI EXPERIMENTAL DESIGN AND GENETIC ALGORITHM

Abstract As genetic algorithm parameters vary depending on different problem types when applying genetic algorithm to reach global optimum, appropriate design value selection has significant impact on the efficiency of genetic algorithm. However, most users adjust parameters manually based on the reference values of previous literature. Such trial-and-error method is time-consuming, ineffective, and often it could not locate the optimal combination. Therefore, in flowshop scheduling problems, this research anticipates to complete optimal parameter combination design in genetic algorithm using Taguchi experimental design. According to the research results, different ways of producing initial solution have significant influence on this research topic. Consequently, confirmation experiment is conducted using the optimal parameter combination obtained from the research results. It is found that the predicted value of signal-to-noise ratio (S/N ratio) and its actual value exists deviation of 0.238%, indicating repetitiveness and robustness of the obtained parameter combination. Hence, this research method can effectively reduce time spent on parameter design using genetic algorithm and increase efficiency of algorithm.

A study of applying data mining to early intervention for developmentally-delayed children

The implementation of early intervention has close relation to the growth development of developmentally-delayed children. The earlier the intervention is involved the more significant effects and results it will bring to the benefits of these young children. However, providing early intervention exclusively without finding out the relationship between the two would eventually leave the problematic point remain unsolved. Since a childs becoming developmental delay is resulted from many factors, much of the valuable knowledge among all needs to be unveiled. In the process of knowledge discovery, use data mining approach to nugget out potential knowledge from vast amounts of data. The main purpose of this study is to explore the hidden knowledge among medical history data of developmentally-delayed children. Fields of medical history database belongs to set and binary, so decision tree is constructed to classify delay levels of each type according to physical illness, and association rule is applied to locate correlations between cognitive, language, motor, and social emotional developmental delays. The study results indicate that the majority of illnesses will result in delays in cognitive, language, and motor development. Simultaneously, among all types of delay, motor and cognitive delay mostly accompanies with symptoms of language delay. The results of this study enable healthcare professionals to be on top of the developments of young children during the process of evaluation and diagnosis, and to provide early intervention so that developmentally-delayed children can catch up with their normal peers in development and growth.

Using case-based reasoning to establish a continuing care information system of discharge planning

As family structure changes, population is aging and disease styles tend to be chronic, long-term care in Taiwan becomes problematic, needs to be addressed, and requires immediate solution. Presently, many medical care institutions in the country have assisted patients in discharge planning; however, the lack of a standard evaluation procedure in the process of discharge planning is disturbing for each hospital. Without it, there might be huge differences in the evaluation results. Moreover, the lack of support and the uncertainty of case eligibility standard in these institutions further affect the performance of continuing care services. This research adopted Case-Based Reasoning to establish a continuing care information system of discharge planning. With previously evaluated information of past cases, the similarity index is compared among new cases. In coordinate with Analytic Hierarchy Process, index weight is calculated to reason an old case that is most closely related to the condition of the new case. This information system can assist discharge-planning staff in accurately formulating a plan of action based on previous case-assessment experience and in obtaining valuable information that helps make decision. Through the implementation of the system, accumulation on knowledge and experience of continuing care models will help staff evaluate process of discharge planning to achieve a reasonable, standardized, and simplified procedure as a whole. This research will transform the evaluating experience of discharge-planning professionals into an assessment method with the application of computer reasoning to make the evaluation process of discharge planning convenient as well as to save more time for discharge-planning professionals to further understand the actual conditions of each case. On the other hand, this information system will provide discharge-planning staff with a set of recommendations as references for making individual discharge plan. It is expected through this research that each hospital be provided with a blue print of improvement in case evaluation process and management.

Quadratic loss functions and signal-to-noise ratios for a bivariate response

Abstract In this paper, the quadratic quality loss functions and signal-to-noise (S/N) ratios for a bivariate response are developed. Quality characteristics by variables are divided into three types: (1) Smaller the better (STB), (2) Larger the better (LTB), and (3) Nominal the best (NTB). The focus is on the bivariate quality characteristic response combinations (NTB, NTB), (STB, STB), and (LTB, LTB) cases. The treatment of mixed bivariate (STB, LTB), (STB, NTB), and (LTB, NTB) response cases will be forthcoming as a sequel to this paper in the next issue of this journal (Vol. 20, No. 2, 2001). The relationships among quality loss constants are also discussed in each case. An example of a robust parameter design experiment using simulated data is provided to illustrate the use of signal-to-noise ratios, which were developed to identify the optimal factor settings. The design matrix used in the example was an L 8 (2 7 ) Taguchi orthogonal array (OA) as a one-half fraction of a 2 4 full factorial experiment.

Using case-based reasoning to diagnostic screening of children with developmental delay

According to the statistics of health organizations of the United Nations, population of children with developmental delay approximately takes 6-9% of the total population. Take the number of newborns (9420) in rural Yunlin County of Taiwan in year 2002 for estimation; there will be approximately 700 (cases regarded developmental delay with suspicion) suspect cases each year. When symptoms of these children are detected early and early intervention given appropriately, wastes in medical resources and social costs could be effectively reduced. Overseas researches show that for children with developmental delay, the best time to intervene is prior to the age of six, and it is golden treatment period before three and a half years old. Studies indicate that when developmentally delayed children receive early intervention, they will show significant improvement in symptoms; some might even recover completely. This study adopts the characteristics of case-based reasoning (CBR) to enhance the screening efficiency of children with developmental delay. CBR is a technology that resolves problems; it resolves currently encountered problems based on previous experiences, which is very similar to the way human beings solve problems by learning from experiences. Since CBR possesses memory functions, which allow it to make judgments and comparisons on new cases, based on old cases saved in the system previously, it is appropriate to apply CBR to a supporting system that has characteristics of invisibility and variation. Therefore, this study uses CBR to establish a screening system of developmental delayed children hoping to increase the screening efficiency. After system verification, the reasoning mean of case similarity is 0.92, and that of accuracy, 0.91, indicating a high level of verification results of this system, and thereby verifying a high level of feasibility of the system.

Application of quality function deployment launches to enhancing nursing home service quality

Abstract Lately, thanks to the prosperity brought by economic development, medical standards are progressing and people are living longer than before. The average lifespan in Taiwan increased to 74.6 years in 1988, an increase of 0.3 years since 1993; simultaneously however, the population of 65 years old or older in Taiwan is rapidly increasing, reaching 1,990,000 in 2002, representing 8.9% of the total population, making Taiwan an ageing society according to definition set out by the World Health Organization (WHO). In the Yunlin County of Taiwan, the proportion of the population over 65 years was 10.8% in 1998; the ageing index is 53.5, which ranked a high fifth among the Taiwan area (the highest is in Penghu County, 70.9, while the lowest is in Taipei County, 27.8). A highly aged population is associated with large-scale health problems, and thus, the need for nursing homes or organizations is rapidly growing. However, the quality of existing nursing homes/institutes is variable. Notably, the main purpose of caring for elderly people in such homes is not only to help elderly people to live long and healthy lives, but also to improve their quality of life and level of dignity. Therefore, this study focuses on a long term nursing home institute located in Yunlin area, examining its service quality and seeking the voice of customers (VOC) by using Quality Function Deployment (QFD), and using the information obtained to assist the organization in improving service quality, weighing costs against benefits, and assigning a weighting to improve service quality; moreover, this study can provide other nursing organizations with a reference basis in promoting service quality. Regarding Quality Function Deployment, this study introduces the concept of fuzzy theory into the calculations used to determine the house of quality. Furthermore, the Fuzzy Analytic Hierarchy Process (FAHP) is introduced to calculate quality based on client requirements. Finally, the fuzzy utility value computation method is used to determine the technical importance. Since the traditional Quality Function Deployment is too subjective and does not consider correlation among quality factors, this breakthrough of adopting the fuzzy utility value computation to determine the technical importance will make the findings of this study much more objective and tending to consider human nature. According to the findings of this study, the first two areas requiring improvement in nursing organizations are the speed to handle the emergency, and the recruitment of specialized medical personnel. Additionally, this study also creates short-term and long-term improvement plans, and allows organizations to use these as a reference in enhancing service quality.

On the bootstrap confidence intervals of the process incapability index CPP

The process incapability index Cpp is an indicator, introduced by Greenwich and Jahr-Schaffrath, for evaluating the capability of a process. When Cpp is applied to evaluate a process, estimating the confidence interval of Cpp is important for statistical inference on the process. Calculating the confidence interval for a process index usually needs the assumption about the underlying distribution. Bootstrapping is a non-parametric, but computer intensive, estimation method. In the present paper we report the results of a simulation study on the behavior of four 95% bootstrap confidence intervals (i.e. standard bootstrap, percentile bootstrap, biased-corrected percentile bootstrap, and biased-corrected and accelerated bootstrap) for estimating Cpp when data are from a specific Burr distribution, which is used to represent various probability distributions. A detailed discussion of the simulation results is presented and some conclusions are provided.

Bivariate tolerance design for lock wheels by considering quality loss

In the design of tolerance allocation the cost–tolerance function is usually employed to represent the objective function which is to be minimized. The traditional cost–tolerance functions in the literature are concerned with only one characteristic. In this paper we obtain a bivariate cost–tolerance function to describe the relationship between the cost and tolerances of two characteristics (i.e. the thickness and inner diameter) of a lock wheel. Then the bivariate loss function is combined with the bivariate cost–tolerance function to determine the optimal tolerances for the thickness and inner diameter of a lock wheel such that the users potential loss/cost may be evaluated. When the quality loss is considered, the tolerances of both characteristics become tighter. By including the effect of product degradation, the present work of expected bivariate quality loss is then introduced as a quality performance measure. By assuming linear drifts on both the thickness and inner diameter of the lock wheels, the model with the present worth of quality loss leads to tighter tolerances of both characteristics. In addition, a longer planning horizon (or a longer useful life of the product) leads to tighter tolerances and a larger users discount rate results in looser tolerances for both characteristics. Copyright

MINIMUM-LOSS DESIGN OF X-BAR CONTROL CHARTS FOR NON-NORMALLY CORRELATED DATA

ABSTRACT When the [xbar] chart is applied to monitor a production process, three parameters should be determined: the sample size, the sampling interval between successive subgroups, and the control limits for the chart. In 1956, Duncan [5] presented the first cost model to determine the three parameters for the X charts, which is called the economic design of [xbar] charts. Alexander et al. [1] combined Duncans cost model with the Taguchi loss function to present a loss model for determining the three parameters. Traditionally, when conducting the design of control charts, one usually assumes the measurements within a subgroup are independently and normally distributed. However, this assumption may not be tenable. In this paper, we develop the minimum-loss design of [xbar] charts for non-normally correlated data, where Alexanders loss model is used as the objective function. An example is provided to illustrate the solution procedure. A sensitivity analysis is performed to investigate the effects of non-normality and correlation coefficient on the optimal design of the chart.