Product Prediction of Mushroom Agricultural Plants Using Machine Learning Techniques

This paper presents mushroom prediction using machine learning methods to identify poisonous and non-toxic mushrooms and find the level of accuracy of several machine learning algorithms. Several algorithms are tested to get the best performance, namely by using Neural Network algorithm, Logistic Regression, Support Vector Machine Learning, Naïve Bayes, Decision Tree, Random Forest. The Neural Network algorithm occupies the accuracy value with the highest performance, with an accuracy value of 92.98%. Next, DDN is optimizing by using RMSprop, Adam SGD, Adagrad, and Adadelta with a learning rate comparison of 20 epochs. The experiment show that it produces better accuracy value, so we conduct some experiments by using Deep Neural Network (DNN) in terms of accuracy value. Deep Neural Network with non-transfer learning produces 99.38% accuracy with the Adagrad optimizer for classification of mushroom plants. The experimental results show that the classification using the Deep Neural Network (DNN) is able to achieve higher accuracy than other algorithms to classify mushroom plants.