Extended Decision Tree for Finding the Top Item sets in Retail Stores
Abstract
Many of the frequent pattern algorithms to find the relation between the items utilizes unsupervised approaches. These algorithms need to generate a huge number of rules with searching criteria for many parameters which results in high results in low performance. To solve this problem, the system constructs an augmented dataset and converts it into a supervised approach based on the items purchased by the users in different localities. To perform this task, the system has tuned the parameters associated with the decision tree. One of the important estimators of the tree is a criterion, the Gini index which determines the impurity of the node. There are many traditional algorithms like SVM, and KNN which can be tuned but when compared to all these approaches including the stacking approach, the decision tree with the best parameters has achieved good accuracy with “99.5%”