Comparative Analysis of Machine Learning Model Performance for Classification of Edible or Non-edible Mushrooms

Abstract

Mushrooms provide significant nutritional benefits and play a crucial role in the global food industry. However, not all mushroom species are safe for consumption, as some contain toxic compounds that can cause severe poisoning and even death. Accurate identification is essential to differentiate between edible and poisonous mushrooms. Traditional classification methods relying on manual morphological identification are often inaccurate, especially when toxic and edible mushrooms have similar physical characteristics. Machine Learning (ML) technology offers an innovative solution to enhance classification accuracy and improve safety in mushroom consumption. This study compares the performance of three major classification algorithms—Random Forest, Logistic Regression, and Naïve Bayes—using an open dataset from Kaggle. The analysis was conducted using the KNIME platform, evaluating the algorithms based on accuracy, sensitivity, and computational efficiency. The results indicate that Random Forest achieved the highest accuracy at 98.90%, followed by Logistic Regression at 69.67% and Naïve Bayes at 55.46%. These findings highlight the superiority of ensemble methods in classification tasks. This research contributes to the development of a reliable ML-based mushroom classification system. However, limitations remain, such as the exclusion of other high-performance algorithms like Support Vector Machine and Artificial Neural Networks. Future studies may incorporate optimization techniques to improve model performance. Additionally, implementing this classification system into mobile or web-based applications could provide broader benefits by enabling quick identification of mushrooms, minimizing health risks, and improving consumer confidence in mushroom safety.