Efficient decision making of data by using Deep learning via generating Chernoff Face

Abstract

In the current era in which we are living data is very important in every field of life. Data allows associations to quantify the adequacy of a given technique, When systems are instituted to conquer a test, collecting data will allow you to decide how well your good answer is performing, and your methodology should be changed or changed over the long haul. Therefore, for a better understanding and decision, we need well-arranged data. However, unfortunately, we do not get it initially. For this purpose, we need to visualize the data in such a way that we can classify between different classes. Until we do not properly classify data, we will not be able to make a good decision based on data. We have different visualization (like scatter plot, histogram, Chernoff faces, Pixel-oriented visualization, Geometric Projection Visualization, Icon-Based Visualization, Hierarchical Visualization, Visualizing Complex Data, and Relations, etc.) and classification (SVM, Decision tree, random forest, Naive Bayes, etc.) approaches for numeric data[1]. We introduce a novel approach for numeric data to increases the efficiency of the data to do a better understanding. Our approach is a combination of the visualization technique Chernoff faces[2] and the CNN model, which is used for the classification of images. For the validation of our approach, we use critical medical data hepatitis C Virus (HCV). This is very important according to the patient perspective. HCV is growing very fast worldwide. And it is the major global cause of death. It is not very dangerous in the early stages but can be deadly later on. The cause of HCV differs from country to country. In Pakistan, the key factor of HCV is the reuse of glass syringes. In Europe and America, the cause of HCV is the use of unsafe drug injections. Classification in this type of dataset is very important because it can save a life. Without using our approach, the efficiency of the data was about 22% to 26% and after using our proposed approaches the efficiency of data to increase up to 99%.