Publication

Abstract : Any study on dataset and inadequate representation are featured in the conventional data mining process. Consequently, data mining findings reveal high chances of challenges and huge root-mean-square approximation errors besides long utilization. The proposed algorithm for associating distribution patterns in streams of medical data is based on density calculations. The distance approach will be utilized to identify any orphaned or questionable data within the data stream, and the gathering of medical data will be used to achieve the aim. Data stream duplicates are compiled by sector match quality, and both the out-of-the-ordinary and the redundant data are eventually eliminated. The density of the information stream is then calculated based on histogram calculations. This investigates the flow of medical information from the viewpoints of concentration and dispersion, as well as the physical characteristics of information distribution, in combination with the results of data density estimate. The model is then built upon a neural network composite, the data distribution parameters in the clustering layer of the model are entered, and in-depth testing is carried out via the neural B.P. network (Back Propagation) on the mining layer of the model. Temporarily, all laws under the grouping of the hidden layer operation value besides the resulting output value are derived, and all rules under the grouping of the hidden layer value besides the conforming contribution value in the medical information stream pattern. The new findings show a curve of contour contiguous to the true curve of the likelihood density, a fair range of medical data dispersion and high medical data consistency, a lower probability of data redundancy, a low RMSEA for mining, a lower time for data recovery.