Deep Associative Classifier

Deep neural networks architecture provides a powerful technique for solving various problems including classification. They owe their performance to the complex and layered data representation and processing built upon neural networks. The success of deep neural networks in various fields has resulted in less focus on other techniques like rule-based models, especially associative classifiers. Associative classifiers are competitive models to deep neural networks on tabular data but suffer from certain limitations i.e., require proper threshold values that differ for different datasets. Even though deep neural networks have resulted in huge success, they have complex and lengthy hyper-parameter tuning. In recent years, attempts to develop models that can compete with deep neural networks using deep representations with decision trees while reducing hyper-parameters have been tried. In this study, we propose a Deep Associative Classifier (DAC), an ensemble of associative classifiers that transforms features in a deep model representation. This model has deep neural network like architecture with associative classifiers as a base learner and overcomes some of the limitations of deep neural network architecture as well as associative classifiers. We use 10 UCI datasets and compare our approach with other existing deep neural network models, a gcForest approach, and associative classifiers. Our proposed model outperforms various state-of-the-art classifiers not only in terms of accuracy but also memory requirement and has fewer hyper-parameters to tune.