هضاب خالد عبيس; نبيل هاشم الاعرجي; ايمان صالح الشمري
المؤلفون بالإنجليزي
;Hadab Khalid Obayes; Nabeel Al – A'araji; EmanAL-Shamery
الملخص العربي
Drug consumption data needs to be linked to the disease. The process of analyzing quantities consumed based on drug name and brand is complex. It needs to be accurate because it is involved in the provision, manufacture,and marketing of medicines. The aim of this paper is to obtain optimal clustersof the drugs according to utilization. Anew model is proposed for the clustering process, specifically the Disease_Drugs_Clustering_Deep_Nural_network (DDC_DNN) as a type of deep neural network. This model consists of four layers. In the first layer,the features have been adapted to the network weights. The normalization and standardization are satisfied in the second layer. The main contributions are concerned in the forming primal clusters according to neighbors' proximity and distance. In the third layer, the final clustering is organized by re-forming clusters depends on the calculation of cluster centers and merging of the nearest clusters according to a carefully selected threshold. Three diseases have been linked with their drugs to be the research data set (diabetes, leukemia,and allergy). The final clusters are optimal clusters. Silhouette validity score has been used to validate the quality of clusters. The result of the proposed model has been compared with the traditional method K-means. Silhouette score of the proposed model result was better than the result of the K-means for the data set
الملخص الانجليزي
Drug consumption data needs to be linked to the disease. The process of analyzing quantities consumed based on drug name and brand is complex. It needs to be accurate because it is involved in the provision, manufacture,and marketing of medicines. The aim of this paper is to obtain optimal clustersof the drugs according to utilization. Anew model is proposed for the clustering process, specifically the Disease_Drugs_Clustering_Deep_Nural_network (DDC_DNN) as a type of deep neural network. This model consists of four layers. In the first layer,the features have been adapted to the network weights. The normalization and standardization are satisfied in the second layer. The main contributions are concerned in the forming primal clusters according to neighbors' proximity and distance. In the third layer, the final clustering is organized by re-forming clusters depends on the calculation of cluster centers and merging of the nearest clusters according to a carefully selected threshold. Three diseases have been linked with their drugs to be the research data set (diabetes, leukemia,and allergy). The final clusters are optimal clusters. Silhouette validity score has been used to validate the quality of clusters. The result of the proposed model has been compared with the traditional method K-means. Silhouette score of the proposed model result was better than the result of the K-means for the data set
هضاب خالد عبيس الجبوري | Hadab khalid obayes | 3744