Preserve Quality Medical Drug Data toward Meaningful Data Lake by Cluster
Year : 2019-09-01
Faculty : Information Technology
Author : محمد احمد محمد الفيومي / عايش منور هويشل الحروب /
Abstarct :
Big data is facing many challenges in different aspects, which appear in characteristics such as: Velocity, Volume, Value and Veracity. Processing and analysis of big data are challenging issues to acquire quality information in order to support an accurate medical drug practice. The quality of data taxonomy is indicated by three basic elements: meaningful, prediction and decision-making. These elements have been encouraged in previous work that focused on the same challenges on big data. Consequently, the proposed approach preserves the quality of medical drug data toward a meaningful data lake by clustering. It consists of four components. Data collection and pre-processing represent the first component in the data lake. Profile data is treated with semi-structured data to clean it up. The second component is extracting data through enforcing rules on whole data to produce different groups and generate weight based on constraints within groups. In component three, data is organized and clustering. This component complies with schema profiling referring to component two in the data lake. Weight outputs of component three are inputs for component four, where K-Mean clustering is applied to obtain different clusters. Each cluster presents an alternative drug to achieve meaningful drug data that is consistent with component three in the data lake. This paper addressed two main challenges; the first challenge is extracting meaningful data from big data, whereas the second challenge is using the big data technique with K-Mean clustering algorithm. An experimental approach was followed by using Food and Drug Administration (FDA) data and symptoms in R framework. ANOVA statistical test was carried out to calculate the sum of square error, P- Value and F-Value for the evaluation of variances between clusters and variances within clusters. The results showed the efficiency of the proposed approach.
Year : 2019-09-01
Faculty : Information Technology
Author : محمد احمد محمد الفيومي / عايش منور هويشل الحروب /
Abstarct :
Big data is facing many challenges in different aspects, which appear in characteristics such as: Velocity, Volume, Value and Veracity. Processing and analysis of big data are challenging issues to acquire quality information in order to support an accurate medical drug practice. The quality of data taxonomy is indicated by three basic elements: meaningful, prediction and decision-making. These elements have been encouraged in previous work that focused on the same challenges on big data. Consequently, the proposed approach preserves the quality of medical drug data toward a meaningful data lake by clustering. It consists of four components. Data collection and pre-processing represent the first component in the data lake. Profile data is treated with semi-structured data to clean it up. The second component is extracting data through enforcing rules on whole data to produce different groups and generate weight based on constraints within groups. In component three, data is organized and clustering. This component complies with schema profiling referring to component two in the data lake. Weight outputs of component three are inputs for component four, where K-Mean clustering is applied to obtain different clusters. Each cluster presents an alternative drug to achieve meaningful drug data that is consistent with component three in the data lake. This paper addressed two main challenges; the first challenge is extracting meaningful data from big data, whereas the second challenge is using the big data technique with K-Mean clustering algorithm. An experimental approach was followed by using Food and Drug Administration (FDA) data and symptoms in R framework. ANOVA statistical test was carried out to calculate the sum of square error, P- Value and F-Value for the evaluation of variances between clusters and variances within clusters. The results showed the efficiency of the proposed approach.