The paper introduces the alternating decision tree (ADTree) learning algorithm, a novel approach that combines decision trees with boosting techniques. A decision tree is a classification rule that maps instances to classes, but it can be large and complex, making it difficult to interpret. AADT generalizes both decision trees and voted decision trees, providing a more interpretable representation. The AADT classifier is represented as a weighted vote of simple prediction rules, making it easy to apply boosting algorithms like AdaBoost. The learning algorithm grows the tree iteratively, adding one base rule at a time, with each rule corresponding to a subtree with a decision node and two prediction nodes. The final classification is the sign of the sum of all base rule predictions. The paper demonstrates that AADT classifiers are competitive with boosted decision tree algorithms like C5.0 but are generally smaller and easier to interpret. Additionally, AADT provides a natural measure of classification confidence, which can be used to improve accuracy by abstaining from predicting hard examples. Experimental results on various datasets show that AADT performs similarly to C5.0 with boosting and offers better interpretability and robustness.The paper introduces the alternating decision tree (ADTree) learning algorithm, a novel approach that combines decision trees with boosting techniques. A decision tree is a classification rule that maps instances to classes, but it can be large and complex, making it difficult to interpret. AADT generalizes both decision trees and voted decision trees, providing a more interpretable representation. The AADT classifier is represented as a weighted vote of simple prediction rules, making it easy to apply boosting algorithms like AdaBoost. The learning algorithm grows the tree iteratively, adding one base rule at a time, with each rule corresponding to a subtree with a decision node and two prediction nodes. The final classification is the sign of the sum of all base rule predictions. The paper demonstrates that AADT classifiers are competitive with boosted decision tree algorithms like C5.0 but are generally smaller and easier to interpret. Additionally, AADT provides a natural measure of classification confidence, which can be used to improve accuracy by abstaining from predicting hard examples. Experimental results on various datasets show that AADT performs similarly to C5.0 with boosting and offers better interpretability and robustness.