Seybold Report ISSN: 1533-9211

Abstract

LUNG DISEASE PREDICTION WITH SPARK FRAMEWORK USING OPTIMAL MODULAR NEURAL NETWORK IN BIG DATA


V Durga Devi
Research Scholar, Department of Computer Applications, VISTAS, Pallavaram, vdurga2216@gmail.com

R Priya
Professor, Department of Computer Applications, VISTAS, Pallavaram, priyaa.research@gmail.com



Vol 17, No 10 ( 2022 )   |  DOI: 10.5281/zenodo.7157438   |   Licensing: CC 4.0   |   Pg no:1674-1689   |   Published on: 07-10-2022



Abstract
The advancements in big data analytics helps healthcare professional in identifying high-risk patients and informed them the status of disease to the patients on time. Moreover, doctors can provide efficient treatment and the cost of extending improved care is also reduced. The major cause for chronic lung disease is smoking. Tobacco smoke either in the form of solid, liquid or gas contains particles comprising of thousands of chemical components which include several toxins and carcinogens. In big data, Artificial Intelligence (AI) is a promising tool which helps in predicting the disease. Under the artificial intelligence the role of Convolution Neural Network (CNN) has many benefits; thus allied in small as well as large scale datasets. However, it is computationally more expensive while the original data is trained which also consumes more time for a complex model even when most powerful GPU hardware is employed. To overcome this issue, this paper proposes MobileNet V2 with Modular Neural Network (MobileNet V2-MNN) for analyzing big data. This model comprises of feature selection and classification stages. During feature selection, the significant features are selected using grasshopper optimization algorithm which consumes very less time. During classification, Neural Network is constructed which uses these selected features for classifying patients resulting in improved accuracy while detecting lung disease; moreover false positive rate is also minimized. This goal is consistently achieved and thus the proposed model outperforms few standard approaches which are considered like DeepRisk, Optimized Dual Attention Neural Network (ODANN), Recurrent Convolutional Neural Network (RCNN). It is observed that the proposed MobileNet V2-MNN achieves 97.1% of accuracy, 96.2% of precision, 89.8% of recall,84.82% of F1-score, 54.66% of AUC and 45.92% of ROC.


Keywords:
Big data, lung disease, neural network, preprocessing, spark, optimization, classification.



Download Full Article PDF


Back to Current Issue Page