An enhanced hybrid colonial competitive rule-based classifier achieved consistently higher accuracy and outperformed both its original formulation and several alternative classification models across nine real-world datasets.
Classification algorithms are widely applied across diverse domains and are commonly grouped into categories including instance-based, neural network-based, and rule-based classifiers. The Colonial Competitive Rule-based classifier operates as a rule-based model utilizing the Imperialist Competitive Algorithm as its core computational engine, extracting rules through the simultaneous evaluation of all dataset features. While the baseline classifier demonstrated promising performance against several standard models in earlier investigations, researchers developed the hybrid framework to further elevate classification accuracy.
The hybrid approach introduces two key methodological contributions. First, rather than beginning execution with randomly initialized rules, the model is initialized using rules generated by the Repeated Incremental Pruning to Produce Error Reduction algorithm, directly enhancing initial rule quality.
Second, the methodology incorporates a fuzzy Mamdani inference system to dynamically adjust a crucial hyperparameter of the Colonial Competitive Rule-based classifier over time, thereby strengthening adaptability. In comparative evaluations across nine real-world datasets, the resulting hybrid architecture consistently exceeded the accuracy of the original classifier as well as other evaluated classification models.
The methodological design of the Colonial Competitive Rule-based classifier centers on using the Imperialist Competitive Algorithm as its core engine to discover classification rules. Although earlier investigations established that the standard Colonial Competitive Rule-based classifier achieved promising results when compared against several standard classifiers, the authors introduced the hybrid method to specifically elevate classification accuracy. A central enhancement targets the rule initialization stage. Rather than relying on randomly initialized rules, the updated methodology initializes the classifier with rules produced by the Repeated Incremental Pruning to Produce Error Reduction algorithm. By integrating rules generated through this pruning algorithm, the hybrid architecture directly improves initial rule quality before running the competitive optimization process.
The second primary contribution focuses on dynamic parameter control during model operation. To improve adaptability, the researchers incorporated a fuzzy Mamdani inference system into the architecture. This fuzzy inference mechanism dynamically adjusts one of the crucial hyperparameters of the Colonial Competitive Rule-based classifier over time as the algorithm proceeds. By pairing the improved initial rule quality derived from the pruning algorithm with the continuous adaptation supplied by the fuzzy Mamdani inference system, the hybrid structure addresses two separate stages of the classifier’s computational lifecycle.
To validate the effectiveness of the hybrid approach, the authors performed empirical experiments across nine real-world datasets and analyzed the outcomes against existing classification techniques. Across these nine datasets, the experimental findings demonstrated that the enhanced Colonial Competitive Rule-based classifier consistently achieved better accuracy than the baseline model. Furthermore, the hybrid model outperformed the original Colonial Competitive Rule-based classifier as well as several alternative classification models evaluated in the study. These comparative results indicate that combining rule initialization via error reduction pruning with fuzzy dynamic hyperparameter adjustment provides measurable performance gains across diverse real-world classification benchmarks.
Subject of Research: Hybrid optimization of colonial competitive rule-based classifiers
Article Title: Optimizing colonial competitive rule-based classifier using a hybrid approach
Article References: Piri, F., & Basiri, A. (2026). Optimizing colonial competitive rule-based classifier using a hybrid approach. Neural Computing and Applications, 38(22), Article 894. https://doi.org/10.1007/s00521-026-12592-0
Image Credits: AI Generated
DOI: 10.1007/s00521-026-12592-0
Keywords: classification algorithms, rule-based classifiers, Colonial Competitive Rule-based classifier, Imperialist Competitive Algorithm, Repeated Incremental Pruning to Produce Error Reduction, fuzzy Mamdani inference system, hyperparameter adjustment, initial rule quality, instance-based classifiers, neural network-based classifiers, nine real-world datasets, classification accuracy, algorithm adaptability, simultaneous feature evaluation, hybrid classification methods, rule extraction
News Source: Denise Maddox. (November 16, 2026). Enhanced Colonial Competitive Classifier Achieves Higher Accuracy Across Nine Datasets. Scienmag.



