CFM-BD: A Distributed Rule Induction Algorithm for Building Compact Fuzzy Models in Big Data Classification Problems

Abstract

Interpretability has always been a major concern for fuzzy rule-based classifiers. The usage of human-readable models allows them to explain the reasoning behind their predictions and decisions. However, when it comes to Big Data classification problems, fuzzy rule-based classifiers have not been able to maintain the good trade-off between accuracy and interpretability that has characterized these techniques in non-Big Data environments. The most accurate methods build too complex models composed of a large number of rules and fuzzy sets, while those approaches focusing on interpretability do not provide state-of-the-art discrimination capabilities. In this paper, we propose a new distributed learning algorithm named CFM-BD to construct accurate and compact fuzzy rulebased classification systems for Big Data. This method has been specifically designed from scratch for Big Data problems and does not adapt or extend any existing algorithm. The proposed learning process consists of three stages: 1) pre-processing based on the probability integral transform theorem; 2) rule induction inspired by CHI-BD and Apriori algorithms; 3) rule selection by means of a global evolutionary optimization. We conducted a complete empirical study to test the performance of our approach in terms of accuracy, complexity, and runtime. The results obtained were compared and contrasted with four state-ofthe-art fuzzy classifiers for Big Data (FBDT, FMDT, Chi-Spark-RS, and CHI-BD). According to this study, CFM-BD is able to provide competitive discrimination capabilities using significantly simpler models composed of a few rules of less than 3 antecedents, employing 5 linguistic labels for all variables.

Do we still need fuzzy classifiers for Small Data in the Era of Big Data?

“…According to these results and the experimental study presented in [5], Big Data classifiers followed similar trends on Big and Small Data. While FMDT was the most accurate method, the resulting trees were significantly more complex than the models built by CFM-BD and FBDT.…”

“…Therefore, the result of each dataset was computed as the average of the five partitions. We considered all the open-source fuzzy classifiers available for Big Data so far 6 (CHI-BD [4], Chi-Spark-RS [6], CFM-BD [5], and FBDT/FMDT [3]) and two of the bestperforming fuzzy classifiers for Small Data (FARC-HD [11] and FURIA [2]). Although the models and learning algorithms used by FRBCSs and FDTs are different, the leaves of FDTs can be converted into a set of IF-THEN rules, allowing us to compare the accuracy-interpretability tradeoff of both types of classifiers.…”

“…• CFM-BD [5] (for Big Data): this fuzzy rule induction algorithm was designed for building compact models that maximize the accuracy-interpretability tradeoff. The learning process consists of three sequential stages: 1) Pre-processing and partitioning: the shape and position of the fuzzy sets are adjusted to the real distribution of training data.…”

“…As for fuzzy sets, both FBDT and FMDT applies a partitioning method that adjusts the number of fuzzy sets for each variable based on the training set, which allowed them to use less fuzzy sets than CFM-BD on Small Data. However, we must remark that the experiments in [5] showed that CFM-BD was able to deal with Big Data problems using less fuzzy sets than FBDT and FMDT, which suggests that CFM-BD might achieve competitive classification performance on Small Data with less fuzzy sets.…”

Spark Architecture and Fractional Artificial Bee Colony-Chaotic Fruitfly RideNN for Big data Classification in Internet of Things

“…However, this method failed to reduce the execution time. Mikel Elkano et al [25] developed a compact fuzzy model for constructing the compact and accurate fuzzy based rule classification system in big data. It specifically handled the big data problems using apriori algorithm.…”