Deep learning disease prediction model for use with intelligent robots.

• Cleveland and Statlog" and a breast cancer dataset known as Wisconsin Breast Cancer (WBC) were attained from the UCI data repository are given to data cleaning process. In the first stage, missing value filling and outlier detection are the two phases, in which Spline Interpolation (SI) is exploited for missing value filling, and entropy, correlation is deployed for outlier detection. • The resultant data from data cleaning are given to the feature extraction process, where Principle Component Analysis (PCA) is employed. Further, the extracted features are multiplied with a weight, and then subjected to the classification process. • The resultant feature vector is conveyed to the Deep Belief Network (DBN) framework. Accordingly, the multiplied weight is optimally tuned by Fitness Oriented Dragonfly Optimization, such that the error between the actual and predicted output is minimized. Deep learning applications with robotics contribute to massive challenges that are not addressed in machine learning. The present world is currently suffering from the COVID-19 pandemic, and millions of lives are getting affected every day with extremely high death counts. Early detection of the disease would provide an opportunity for proactive treatment to save lives, which is the primary research objective of this study. The proposed prediction model caters to this objective following a stepwise approach through cleaning, feature extraction, and classification. The cleaning process constitutes the cleaning of missing values ,which is proceeded by outlier detection using the interpolation of splines and entropy-correlation. The cleaned data is then subjected to a feature extraction process using Principle Component Analysis. A Fitness Oriented Dragon Fly algorithm is introduced to select optimal features, and the resultant feature vector is fed into the Deep Belief Network. The overall accuracy of the proposed scheme experimentally evaluated with the traditional state of the art models. The results highlighted the superiority of the proposed model wherein it was observed to be 6.96% better than Firefly, 6.7% better than Particle Swarm Optimization, 6.96% better than Gray Wolf Optimization ad 7.22% better than Dragonfly Algorithm. [ABSTRACT FROM AUTHOR]