Granule structures, distances and measures in neighborhood systems

BGRF: A broad granular random forest algorithm

 The random forest is a combined classification method belonging to ensemble learning. The random forest is also an important machine learning algorithm. The random forest is universally applicable to most data sets. However, the random forest is difficult to deal with uncertain data, resulting in poor classification results. To overcome these shortcomings, a broad granular random forest algorithm is proposed by studying the theory of granular computing and the idea of breadth. First, we granulate the breadth of the relationship between the features of the data sets samples and then form a broad granular vector. In addition, the operation rules of the granular vector are defined, and the granular decision tree model is proposed. Finally, the multiple granular decision tree voting method is adopted to obtain the result of the granular random forest. Some experiments are carried out on several UCI data sets, and the results show that the classification performance of the broad granular random forest algorithm is better than that of the traditional random forest algorithm.

An intrusion detection method based on granular autoencoders

Intrusion detection systems have become one of the important tools for network security due to the frequent attacks brought about by the explosive growth of network traffic. Autoencoder is an unsupervised learning model with a neural network structure. It has a powerful feature learning capability and is effective in intrusion detection. However, its network construction suffers from overfitting and gradient disappearance problems. Traditional granular computing methods have advantages in solving such problems, but the process is relatively complex, the granularity dimension is high, and the computational cost is large, which is not suitable for application in intrusion detection systems. To address these problems, we propose a novel autoencoder: Granular AutoEncoders (GAE). The granulation reference set is constructed by random sampling. The granulation of training samples is based on single-feature similarity in a reference set to form granules. The granulation of multiple features results in granular vectors. Some operations of granules are defined. Furthermore, we propose some granular measures, including granular norms and granular loss functions. The GAE is further applied to the field of intrusion detection by designing an anomaly detection algorithm based on the GAE. The algorithm determines whether the network flows are anomalous by comparing the difference between an input granular vector and its output granular vector that is reconstructed by the GAE. Finally, some experiments are conducted using an intrusion detection dataset, comparing multiple metrics in terms of precision, recall, and F1-Score. The experimental results validate the correctness and effectiveness of the intrusion detection method based on GAE. And contrast experiments show that the proposed method has stronger ability for detecting anomalies than the correlation algorithms.

An LVQ clustering algorithm based on neighborhood granules

Learning Vector Quantization (LVQ) is a clustering method with supervised information, simple structures, and powerful functions. LVQ assumes that the data samples are labeled, and the learning process uses labels to assist clustering. However, the LVQ is sensitive to initial values, resulting in a poor clustering effect. To overcome these shortcomings, a granular LVQ clustering algorithm is proposed by adopting the neighborhood granulation technology and the LVQ. Firstly, the neighborhood granulation is carried out on some features of a sample of the data set, then a neighborhood granular vector is formed. Furthermore, the size and operations of neighborhood granular vectors are defined, and the relative and absolute granular distances between granular vectors are proposed. Finally, these granular distances are proved to be metrics, and a granular LVQ clustering algorithm is designed. Some experiments are tested on several UCI data sets, and the results show that the granular LVQ clustering is better than the traditional LVQ clustering under suitable neighborhood parameters and distance measurement.

The cost-sensitive approximation of neighborhood rough sets and granular layer selection

Neighborhood rough sets (NRS) are the extended model of the classical rough sets. The NRS describe the target concept by upper and lower neighborhood approximation boundaries. However, the method of approximately describing the uncertain target concept with existed neighborhood information granules is not given. To solve this problem, the cost-sensitive approximation model of the NRS is proposed in this paper, and its related properties are analyzed. To obtain the optimal approximation granular layer, the cost-sensitive progressive mechanism is proposed by considering user requirements. The case study shows that the reasonable granular layer and its approximation can be obtained under certain constraints, which is suitable for cost-sensitive application scenarios. The experimental results show that the advantage of the proposed approximation model, moreover, the decision cost of the NRS approximation model will monotonically decrease with granularity being finer.

Feature Selection Combining Information Theory View and Algebraic View in the Neighborhood Decision System

Feature selection is one of the core contents of rough set theory and application. Since the reduction ability and classification performance of many feature selection algorithms based on rough set theory and its extensions are not ideal, this paper proposes a feature selection algorithm that combines the information theory view and algebraic view in the neighborhood decision system. First, the neighborhood relationship in the neighborhood rough set model is used to retain the classification information of continuous data, to study some uncertainty measures of neighborhood information entropy. Second, to fully reflect the decision ability and classification performance of the neighborhood system, the neighborhood credibility and neighborhood coverage are defined and introduced into the neighborhood joint entropy. Third, a feature selection algorithm based on neighborhood joint entropy is designed, which improves the disadvantage that most feature selection algorithms only consider information theory definition or algebraic definition. Finally, experiments and statistical analyses on nine data sets prove that the algorithm can effectively select the optimal feature subset, and the selection result can maintain or improve the classification performance of the data set.