Performance Analysis of Decision Tree Algorithms on Mushroom Dataset

Abstract

Data mining is a process of extracting valuable information from vast dataset databases. This paper discusses data mining algorithms namely ID3, CART, and HoeffdingTree (HT) based on decision tree. Our goal is to assess the risk to human life through the consumption of mushrooms. Decision tree is used to classify the data with 22 attributes, to either edible or poisonous Mushrooms. Decision tree is used in this study as classification technique for analyzing mushroom data set. This experiment has been performed in R studio software environment. This study illustrates the accuracy of each classifier and the results are compared and discussed to which classifier is best for mushroom dataset. Hoeffding Tree provides better results with highest accuracy, low time and least error rate when compared with ID3 and CART.

Edibility Detection of Mushroom Using Ensemble Methods

“…The highest accuracy of [11] is 99.88% for ANFIS. Table 7 indicates that the accuracy of the candidate system is higher than all methods except HOFFDING tree of [10]. The performance comparison of highest accuracy, as well as the lowest error rate of the proposed method with the previously described existing methods, is shown in Fig.6.…”

“…The performance of proposed methods and other existing works is compared based on the performance measures, Accuracy and Error Rate on the mushroom dataset. We have compared the best result of [9][10][11] for all methods used in these papers and our proposed method are shown in Table 7. In [9], they have used three data mining algorithms Bayes Net, Naï ve Bayes, ZeroR and measured performance of the algorithms with the respective training dataset size and their highest accuracy is 97.22% for Bayes Net when 70% tuples are used for the training set and 30% are for testing.…”

“…Similarly, the highest accuracy for Naï ve Bayes, ZeroR is 96.81%, 64.25% for same data size. In [10], they have analyzed the performance of ID3, CART, Hoeffding tree on the mushroom dataset and the accuracy of ID3 is 69%, CART is 90% and Hoeffding tree is 100%. In paper [11], they have used three classifiers: ANN, ANFIS, Naï ve Bayes.…”

“…It shows that the highest accuracy of [9] is 97.22% for Bayes Net to test using 30% data of mushroom dataset among their three algorithms Bayes Net, Naï ve Bayes, and ZeroR. The highest accuracy of [10] is 100% among the three algorithms ID3, CART, HOFFDING tree for testing all data. The highest accuracy of [11] is 99.88% for ANFIS.…”

“…For classifying mushroom, a narrow range of studies has been done using classification. [5,[7][8][9][10][11] are some of them where Decision Tree (C4.5), SVM (Support Vector Machine), ANN, ANFIS Naï ve Bayes, Bayes Net, ZeroR, and RIDOR single classifier algorithms have been used to classify mushroom. But the performance of single classifier algorithms is poorer than ensemble methods.…”