Breast Cancer Prediction Using Deep Learning and Machine Learning Techniques
Breast Cancer Prediction Using Deep Learning and Machine Learning Techniques
复制标题
使用深度学习和机器学习技术预测乳腺癌
DOI：
10.2139/ssrn.3558786
复制
发表时间：
2020
期刊：
影响因子：
--
通讯作者：
Reena Lokare
中科院分区：
文献类型：
--
作者：
Monika Tiwari;Rashi Bharuka;Praditi Shah;Reena Lokare
乳腺癌主要发生在妇女中，是妇女死亡率上升的主要原因。乳腺癌的诊断是耗时的，并且由于系统的可用性较低，因此有必要开发一种可以在早期阶段自动诊断乳腺癌的系统。各种机器学习和深度学习算法已用于良性和恶性肿瘤的分类。使用了威斯康星州乳腺癌数据集，该数据集包含569个样本和30个特征。本文重点介绍了各种模型，如逻辑回归，支持向量机（SVM）和K最近邻（KNN），多层感知器分类器，人工神经网络（ANN）等。这些算法中的每一个都被测量并比较了所获得的准确度和精度。所有的技术都是用Python编写的，并在Google Colab中执行，这是一个科学的Python开发环境。实验表明，支持向量机和随机森林分类器是最好的预测分析，准确率为96.5%。为了提高预测的准确性，已经实现了CNN和ANN等深度学习算法。在ANN和CNN的情况下获得的最大精度分别为99.3%和97.3%。诸如Relu和sigmoid的激活函数已被用于根据概率来预测结果。
Breast Cancer is mostly identified among women and is a major reason for increasing the rate of mortality among women. Diagnosis of breast cancer is time consuming and due to the lesser availability of systems it is necessary to develop a system that can automatically diagnose breast cancer in its early stages. Various Machine Learning and Deep Learning Algorithms have been used for the classification of benign and malignant tumours. The Wisconsin Breast Cancer Dataset has been used which contains 569 samples and 30 features. The paper emphasises on various models that is implemented such as Logistic Regression, Support Vector Machine (SVM) and K Nearest Neighbour (KNN), Multi-Layer perceptron classifier, Artificial Neural Network(ANN)) etc. on the dataset taken from the repository of Kaggle. Each of these algorithms has been measured and compared with respect to accuracy and precision obtained. All the techniques are coded in python and executed in Google Colab, which is a Scientific Python Development Environment. The experiments have shown that SVM and Random Forest Classifier are the best for predictive analysis with an accuracy of 96.5%. To increase the accuracy of prediction, deep learning algorithms such as CNN and ANN have been implemented. The maximum accuracy obtained in the case of ANN and CNN are 99.3% and 97.3% respectively. Activation functions such as Relu and sigmoid have been used to predict the outcomes in terms of probabilities.