Lung Cancer Prediction using Machine Learning

This research proposes a hybrid strategy that combines deep learning and conventional machine learning methods for image classification. Initially, images are preprocessed using Image Data Generator to resize them to 224x224 pixels and normalize pixel values to the range [0, 1]. A pre-trained MobileNet model (excluding the top layer) is utilized for feature extraction. The feature map is then transformed into a flat vector using a Global Average Pooling layer. Various combinations of classifiers are employed for prediction, including SVM + Random Forest, Random Forest + Logistic Regression, SVM + Logistic Regression, and a full ensemble of SVM + Random Forest + Logistic Regression. For aggregating predictions, the ensemble approach uses soft voting based on the average of each classifier’s probabilities. StratifiedKFold cross-validation is used to maintain the proportion of each class during testing, ensuring reliable model evaluation. Model performance is assessed using accuracy, precision, recall, F1-score, and a confusion matrix against test and validation sets to compare the effectiveness of different classifier combinations. Keywords: Image classification, MobileNet, feature extraction, ensemble learning, Support Vector Machine, Random Forest, Logistic Regression, StratifiedKFold, cross-validation, soft voting.