A Novel Twdlnn And Mining Based Breast Cancer Prediction System in A Big Data Environment

Abstract

Amongst women, Breast Cancer (BC) has turned out to be the main reason for mortality. Predicting the BC early can help in saving the women as of the severe stage of cancer. Though most existing research has been utilizing disparate algorithms for prediction, they still lack in some areas, like accurate prediction and the execution speed. Thus, to trounce such cons, this paper proposed a novel Target Weight based Deep Learning Neural Network (TWDLNN) and mining based BC prediction system on a Big Data (BD) environment. The proposed paper totally comprises ‘4’ steps: i) pre-processing, ii) Feature Selection (FS), iii) rule mining, and iv) classification. First, the Hadoop Distributed File Systems (HDFS) Map-Reduce (MR) function removes the redundant data, and also the missing attributes are swapped in the pre-processing step. Then, the Levy Flight based Chickens Swarm Optimizations (LFCSO) selects the vital features. Subsequently, the Associations Rule Mining (ARM) process is executed, wherein the CFI is attained. Next, the closed frequent itemset (CFI) is inputted to the TWDLNN algorithm that classifies the inputted data into a normal or cancer patient. In the experimental investigation, the proposed TWDLNN’s performance is contrasted with the existing DLNN, ANN, SVM, along with RF-centred on the accuracy as well as execution time metrics.