Associative learning on imbalanced environments: An empirical study

Abstract

Associative memories have emerged as a powerful computational neural network model for several pattern classification problems. Like most traditional classifiers, these models assume that the classes share similar prior probabilities. However, in many real-life applications the ratios of prior probabilities between classes are extremely skewed. Although the literature has provided numerous studies that examine the performance degradation of renowned classifiers on different imbalanced scenarios, so far this effect has not been supported by a thorough empirical study in the context of associative memories. In this paper, we fix our attention on the applicability of the associative neural networks to the classification of imbalanced data. The key questions here addressed are whether these models perform better, the same or worse than other popular classifiers, how the level of imbalance affects their performance, and whether distinct resampling strategies produce a different impact on the associative memories. In order to answer these questions and gain further insight into the feasibility and efficiency of the associative memories, a large-scale experimental evaluation with 31 databases, seven classification models and four resampling algorithms is carried out here, along with a non-parametric statistical test to discover any significant differences between each pair of classifiers.

CDSMOTE: class decomposition and synthetic minority class oversampling technique for imbalanced-data classification

“…For further evaluation, we compared our method with recent state-of-the-art techniques using the most recent results and reported the same experiment settings and datasets. First, we consider Cleofas-Sanchez et al [47], who attempted class-imbalance classification using 31 of the datasets through a hybrid associative classifier with translation (HACT) based on SMOTE and used Gmean for evaluating the results. Table 4 lists the performance of CDSMOTE against this method.…”

“…Table 6 compares CDSMOTE performance against BEPILD using the two metrics reported by the authors (AUC and Gmean). Table 4 compares CDSMOTE with [47] in terms of Gmean.Notice that CDSMOTE obtains the better results in 20 out of 31 datasets. Using a paired t test on the 20 datasets where CDSMOTE wins shows a statistically significant difference with p value equal to 0.000506.…”

“…These were chosen as they are among the most common undersampling methods in the literature. Moreover, CDSMOTE is compared against class decomposition [7] and with recent and stateof-the-art methods including [36,47,48]. The following subsections describe the experiment in details.…”

“…A collection of 60 datasets was used in this experiment, and these are publicly available and commonly used in class-imbalance data classification (i.e., [36,47,48], ...). The datasets were obtained from the KEEL repository.…”

Impact of Imbalanced Datasets Preprocessing in the Performance of Associative Classifiers

“…It is possible to find several state-of-the-art articles [48][49][50][51][52] in which the preprocessing of datasets is employed to reduce the impact caused by the distribution of classes. In such research, it has been empirically demonstrated that the application of a preprocessing stage to balance the distribution of classes is usually a useful solution to improve the quality of the identification of new instances.…”