Rule-Based Modeling of Low-Dimensional Data with PCA and Binary Particle Swarm Optimization (BPSO) in ANFIS

Afnan Al-ALi, Uvais Qidwai · 2025

Read Rule-Based Modeling of Low-Dimensional Data with PCA and Binary Particle Swarm Optimization (BPSO) in ANFIS by Afnan Al-ALi, Uvais Qidwai (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2502-03895.txt.

Introduction

Afnan Al-Ali and Uvais Qidwai slim down fuzzy rule systems in this 2025 paper. Fuzzy rules stay readable for people, a plus in health and finance work. But grid-based rule sets explode in size as inputs grow, even for modest data.

Their method runs principal component analysis on rule firing strengths to find lean components. A binary swarm search then picks which components to keep, cutting rules hard while guarding accuracy. A custom update step tunes key layers and avoids poor local fits. Tests used standard datasets plus a real stroke dataset. Results showed fewer rules, faster training, and high accuracy held across tasks.

Catalog record

CREATORAfnan Al-ALi, Uvais Qidwai
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE81 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2502-03895

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Rule-Based Modeling of Low-Dimensional Data with PCA and Binary Particle Swarm Optimization (BPSO) in ANFIS Afnan Al-ALi, Uvais Qidwai Abstract Fuzzy rule-based systems interpret data in low-dimensional domains, providing transparency and interpretability. In contrast, deep learning excels in complex tasks like image and speech recognition but is prone to overfitting in sparse, unstructured, or low-dimensional data. This interpretability is crucial in fields like healthcare and finance. Traditional rule-based systems, especially ANFIS with grid partitioning, suffer from exponential rule growth as dimensionality increases. We propose a strategic rule-reduction model that applies Principal Component Analysis (PCA) on normalized firing strengths to obtain linearly uncorrelated components. Binary Particle Swarm Optimization (BPSO) selectively refines these components, significantly reducing the number of rules while preserving precision in decision-making. A custom parameter update mechanism fine-tunes specific ANFIS layers by dynamically adjusting BPSO parameters, avoiding local minima. We validated our approach on standard UCI respiratory, keel classification/regression datasets, and a real-world ischemic stroke dataset, demonstrating adaptability and practicality. Results indicate fewer rules, shorter training, and high accuracy, underscoring the method’s effectiveness for low-dimensional interpretability and complex data scenarios. This synergy of fuzzy logic and optimization fosters robust solutions. Our method contributes a powerful framework for interpretable AI in multiple domains. It addresses dimensionality, ensuring a rule base. uzzy rule-based systems excel in interpreting data, especially in low-dimensional scenarios. In contrast, deep learning, while effective in areas like image and speech recognition, struggles with sparse, unstructured, or low-dimensional data due to extensive parameter requirements and overfitting risks. Rule-based systems, rooted in fuzzy logic, provide transparency in decision-making, offering clear rationales for each rule. This interpretability is crucial in fields like healthcare and finance. However, these systems face challenges when dealing with complex and high-dimensional data. To address these challenges, one approach is to integrate rule-based systems with other machine-learning methods. An example is the Adaptive Neuro-Fuzzy Inference System (ANFIS), which combines the transparency of fuzzy rule-based systems with the adaptability of neural networks. However, traditional ANFIS has limitations, particularly when using grid partitioning for rule generation. An issue is exponential growth in the rule count as problem dimensionality increases despite its simple implementation. This research introduces an innovative model for strategic rule reduction to address the challenges of rule-based systems, especially ANFIS with grid partitioning. The model incorporates Principal Component Analysis (PCA) on the normalized firing strengths, transforming them into linearly uncorrelated components. These components are selectively optimized and evaluated using Binary Particle Swarm Optimization (BPSO), resulting in a significant rule reduction that enhances decision-making precision, a critical aspect of rule-based systems. Additionally, a custom parameter update mechanism fine-tunes specific ANFIS layers by dynamically adjusting BPSO parameters, thereby avoiding potential issues related to local minima. These enhancements have been validated on standard datasets from UCI respiratory and keel, covering classification and regression tasks, as well as a real-world ischemic stroke dataset, highlighting the adaptability and practicality of our model. The results from our model demonstrated a marked decrease in the number of generated rules and training duration while maintaining a high level of accuracy. [type=editor, auid=000,bioid=1, ] [type=editor, auid=000,bioid=2, ] keywords Adaptive Neuro-Fuzzy Inference System,Binary Particle Swarm Optimization Technique ,Principal Component Analysis,rules reduction,grid partitioning † † corresponding: Corresponding author † † credit: Conceptualization of this study, Methodology, Software 1 Introduction Fuzzy modeling is a descriptive language that a fuzzy logic system uses to represent real-world operations. These models can express expert knowledge through fuzzy if-then rules without requiring a detailed qualitative analysis. As a result, the ambiguity and uncertainty systems can be represented more transparently, enabling a thorough comprehension of the system’s operation [ 1 ] . Fuzzy rule-based systems are particularly adept at interpreting data in scenarios with fewer variables but complex relationships. In contrast, deep learning is renowned for its success in processing high-dimensional data, such as images, videos, and speech in different applications [ 2 ] [ 3 ] [ 4 ] [ 5 ] [ 6 ] [ 7 ] , where it leverages vast amounts of data to uncover intricate patterns. However, deep learning techniques often face challenges when dealing with sparse, unstructured, or low-dimensional datasets due to their high parameter demands and susceptibility to overfitting. These challenges highlight the advantages of fuzzy modeling in providing clear and interpretable rules for systems characterized by uncertainty and ambiguity. In 1965, Lotfi Zadeh introduced Fuzzy Logic as a mathematical tool to deal with uncertainty by using what is called membership functions , such that the linguistic words like heavy, tall, low, etc., which cannot be categorized precisely in terms of “0” or “1”; instead it is possible to express them in degrees of belonging to a specific category. There are two popular types of fuzzy inference techniques: the Mamdani, which Mamdani and Assilian founded [ 8 ] , and the Sugeno or Takagi–Sugeno–Kang or (TSK) fuzzy inference technique, which Sugeno pioneered [ 9 ] . The key distinction between the two approaches is the outcome of fuzzy rules. TSK fuzzy inference systems use linear functions of input variables as rule consequents, whereas Mamdani fuzzy systems use fuzzy sets as rule consequents [ 10 ] . One of the successful fuzzy inference techniques that can deal with highly complex, nonlinear systems is the Adaptive Neuro-Fuzzy Inference system (ANFIS). The typical way to write a single fuzzy rule in ANFIS based on the form of TSK inference for two inputs x_{1} and x_{2} is: IF x_{1} is A_{1} AND x_{2} is A_{2} THEN y =f( x_{1} , x_{2} ), where: A_{1} , and A_{2} are the fuzzy sets for the input variables x_{1} and x_{2} , y is the output, AND is a logical AND gate [ 11 ] . ANFIS offers advantages but also faces specific challenges [ 12 ] . These challenges associated with ANFIS are interconnected, creating a complex landscape for its application. The need for expert knowledge and rule-based development is closely tied to rule-based complexity since managing and maintaining a complex rule base becomes more challenging and time-consuming. The complexity of the rule base also impacts subjectivity and interpretability, as the intricate rules may lead to inconsistencies and subjective interpretations, making it harder to ensure transparency. Furthermore, integrating ANFIS with other computational techniques presents challenges, particularly when dealing with a complex rule base and issues of subjectivity. Scalability and efficiency concerns are magnified when the rule base becomes more complex, especially in large-scale problems where resource management becomes crucial. Finally, efficient data-driven learning in ANFIS is essential to address the challenges of overfitting and model complexity, which are closely related to the complexity of the rule base. These challenges collectively highlight the intricate nature of ANFIS and the need for careful consideration and expertise in its application [ 13 ] [ 14 ] . From the challenges delineated earlier, we can distill these complexities into three primary focal points, which we intend to concentrate on within the context of this research study, as depicted in the illustrative Figure 1 Figure 1: Main Challenges in ANFIS. Addressing these challenges in ANFIS requires domain expertise, careful rule-based development, thoughtful integration with other techniques, and considerations for scalability and interpretability. Additionally, advances in machine learning techniques, methodologies, and domain-specific knowledge can contribute to overcoming these

Read the complete text