APPLYING DEEP LEARNING FOR HEALTHCARE IN SMART CITY VIA INTERNET OF THINGS

Abstract

In the traditional city for healthcare in IoT, it has been proposed to replace traditional yield models with mathematical models that do not require the assumption of defect density functions. The selection of input parameters in these models is very important, and all the variation factors on the wafer must be included as far as possible. The factors of clustering are usually described by clustering indicators, but some specific clustering patterns will cause the clustering indicators to misjudge the clustering degree, resulting in the yield estimation error becoming larger. In view of this, the proposed study has classified the defect patterns on the wafer into four types: random distribution, regional concentrated distribution, linear distribution and circular distribution, by means of three pattern characteristics analysis. A comparison is made only using cluster indicators to describe cluster phenomena and a model that uses cluster indicators and cluster graphs to describe cluster phenomena. The research results show that when constructing the yield model, the clustering pattern and the clustering index are used to describe the clustering phenomenon in smart city via Internet of things, which is preferred to solely considering the clustering index, as the consequent degree of accuracy far exceeds the improvement of changing the “number of effective grains” in relation to the clustering index. Therefore, the yield rate can be estimated more accurately by using clustering graphs with clustering indicators; the estimated yield in the yield model, with the clustering pattern parameter, is indeed closer to the actual yield than the yield model without the clustering pattern parameter.