mushroom_edibility_prediction

The aim of this analysis was to develop a supervised machine learning model to accurately classify of mushrooms as edible or poisonous. Leveraging a dataset comprising over 60,000 mushrooms with 20 distinct features, including physical attributes like cap measurements, stem characteristics, and gill properties, I fit models using logistic regression, linear discriminant analysis (LDA), k-nearest neighbors (kNN), random forests, and boosting algorithms. I assessed model performance using various metrics such as accuracy, precision, F1 score, log loss, confusion matrices, and ROC curves. Results indicate that while logistic regression and LDA models exhibit subpar predictive accuracy, kNN, random forests, and boosting models demonstrate strong performance, particularly in precision and accuracy. Feature importance analysis reveals the critical significance of attributes like stem width, gill attachment, and stem color in predicting mushroom toxicity.