Heuristic Feature Research Articles

Identification of stromal cell proportion-related genes in the breast cancer tumor microenvironment using CorDelSFS feature selection: implications for tumor progression and prognosis.

Background: The tumor microenvironment (TME) of breast cancer (BRCA) is a complex and dynamic micro-ecosystem that influences BRCA occurrence, progression, and prognosis through its cellular and molecular components. However, as the tumor progresses, the dynamic changes of stromal and immune cells in TME become unclear. Objective: The aim of this study was to identify differentially co-expressed genes (DCGs) associated with the proportion of stromal cells in TME of BRCA, to explore the patterns of cell proportion changes, and ultimately, their impact on prognosis. Methods: A new heuristic feature selection strategy (CorDelSFS) was combined with differential co-expression analysis to identify TME-key DCGs. The expression pattern and co-expression network of TME-key DCGs were analyzed across different TMEs. A prognostic model was constructed using six TME-key DCGs, and the correlation between the risk score and the proportion of stromal cells and immune cells in TME was evaluated. Results: TME-key DCGs mimicked the dynamic trend of BRCA TME and formed cell type-specific subnetworks. The IG gene-related subnetwork, plasmablast-specific expression, played a vital role in the BRCA TME through its adaptive immune function and tumor progression inhibition. The prognostic model showed that the risk score was significantly correlated with the proportion of stromal cells and immune cells in TME, and low-risk patients had stronger adaptive immune function. IGKV1D-39 was identified as a novel BRCA prognostic marker specifically expressed in plasmablasts and involved in adaptive immune responses. Conclusions: This study explores the role of proportionate-related genes in the tumor microenvironment using a machine learning approach and provides new insights for discovering the key biological processes in tumor progression and clinical prognosis.

Estimation psychophysiological state via nonlinear dynamic integral models

The method of experimental research “input-output”of the human oculo-motor system was developed and implemented using innovative eye-tracking technology for recording oculo-motor systemresponses to test visual stimuli. Stimuli are displayed on the monitor screen at different distances from the starting position. This formally corresponds to the action of step signals with different amplitudes at the input of the oculo-motor system. According to the empirical data of the “input-output”studies of the respondent's oculo-motor systemobtained with the aid of the TobiiProTX300 eye tracker, the transient functions of the first and diagonal intersections of the transient functions of the second and third orders of the oculo-motor systemwere determined. Experimental studies of the respondent's oculo-motor systemto identify the state of fatigue were carried out before the beginning (in the morning) and after the working day (in the evening). The obtained multidimensional transient functions are used as a source of primary data in the implementation of intelligent information technology for diagnosis and monitoring of the psychophysiological state of a person. Instrumental algorithmic and software tools for determining diagnostic features based on the identification data of the oculo-motor systemin the form of multidimensional transient functionsin the Python language have been developed.Training samples of data fortwo states of the respondent (“Normal”and “Fatigue”) were formed on the basis of the proposed heuristic features, which are determined using integral and differential transformations of the obtained multidimensional transient functionsof the oculo-motor system. Training samples of data are used to build classifiers of psychophysiological states of an individual using machine learning tools. The informativeness of individual features and all their possible combinations in pairs according to the indicator of the probability of correct recognition was studied using the method of complete search. The research results were obtained by evaluating the quality of recognition of states built by Bayesian classifiers in different spaces of the proposed features. An analysis of the stability of the correct recognitioninformativeness indicator of different feature spaces under the influence of different levels of additive noise on the features was carried out. Two-dimensional feature spaces with the maximum and most stable value of the correct recognitionindicator were found when solving the scientific and practical task of assessing the psychophysiological state (fatigue) of a person (0.9375). Thus, it seems appropriate to use the multidimensional transient functionsobtained from eye-tracking data in diagnostic studies in the fields of neuroscience and experimental psychology

Feature Selection Approach Based on Improved Fuzzy C-Means With Principle of Refined Justifiable Granularity

Fuzzy C-Means (FCM) is a clustering algorithm based on partition of the universe. However, the partition generated by an equivalence relation is strict in practical application and exhibits relatively poor fault-tolerant mechanism. In this paper, a novel binary relation based on improved FCM with the principle of refined justifiable granularity is presented. Different expressions of the proposed binary relation under different values of weight parameter are discussed, and the changes of the properties of the binary relation under different parameter values are provided. By measuring the significance of attributes in the feature space, a feature selection method, called forward heuristic feature selection (FHFS), is designed to construct the low-dimension feature space based on maximizing the original data and information retention through the defined degrees of aggregation and dispersion. It is shown how the results of feature selection and classification performance vary when the values of the weight factor locate in different ranges. To illustrate the superiority and effectiveness of the proposed FHFS algorithm, nine high-dimensional datasets and eight image datasets from UCI repository are used and compared with other feature selection methods, respectively. The results of experimental evaluation and the significance test show that the proposed learning mechanism is a superior algorithm.

Privacy Preserving Contraband Detection Using a Millimeter-Wave Dynamic Antenna Array

We present a privacy preserving technique for contraband detection using a real-time 75 GHz rotational dynamic antenna array that samples only a reduced set of Fourier-domain information to prevent image reconstruction of the screened subject. Complex objects such as handguns have sharp edges that correspond to sharp spatial frequency features in the Fourier domain. These features are dependent on the orientation of the object and manifest at millimeter-wave frequencies through conditions such as being concealed under clothing. The technique combines a dynamically rotated active interferometer and simple arithmetic feature extraction approach supporting a threshold-based classifier to identify when a handgun is concealed under clothing. The active interferometer technique utilizes noise transmitters to illuminate the test subject to produce thermal-like radiation, enabling Fourier-domain signal sampling, while the rotational dynamics support sampling that spans a circle in the 2-D Fourier domain to retrieve sharp-edge responses induced by concealed objects. We experimentally demonstrate the detection of concealed objects under clothing in a laboratory environment using a 75 GHz dynamic antenna array based on simple heuristic features extracted from statistics of the measured responses. The classifier is evaluated based on processing one or more subsequent measurements, showing a classification accuracy of 0.908 and <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$F$</tex-math> </inline-formula> 1-score of 0.916 using four consecutive measurements.

Exploring network reliability by predicting link status based on simplex neural network

Complex networks are graph-based structures with non-trivial topological features that frequently occur in real systems. Link prediction plays an important role in various real-world networks application, such as recommendation systems, protein structure prediction, packet forwarding strategy optimization, etc. The existing link prediction approaches mainly focus on superficial heuristic features, while ignoring high-order structure information. In this paper, we propose a deep-learning based model, named Weisfeiler–Lehman Simplex Neural Network (WL-SNN), which can learn the high-order simplex information of the network. In particular, we design a third-order Laplace operator to extract the simplicial features and utilize the graph convolutional network to compensate for the possible deficiencies of the model resulting from the single-channel features. Furthermore, we use the Weisfeiler–Lehman algorithm to extract closed subgraphs of the target, which significantly enhances the adaptability of the model to large-scale networks. Experimental results on six real-world networks show that our approach achieves comparable performance in the link prediction task as well as in the stability analysis of the network.

A self-adaptive local diversity-preserving NSGA-II for the multidisciplinary design optimization of a miniaturized supersonic rocket

The miniaturized supersonic rocket or missile is an innovative concept most notably characterized by its small caliber, short length, and high flight speed. Tightly coupled design parameters and constrained objectives bring high complexity to launch safety assessments. The commonly used heuristic algorithm NSGA-II cannot address this complexity with its original density estimation strategy. In this paper, a multidisciplinary design optimization (MDO) based on a multidisciplinary feasible (MDF) architecture is established and simplified according to specific design requirements. The design system includes one coupled numerical model, two drivers, and six disciplines. As an optimizer for an MDO problem, multi-objective evolutionary algorithms (MOEAs) with heuristic features are very conducive to exploring the performance of complex rocket systems. The coupling relationship between the design variables of a small dual-chamber rocket is complex, and its optimal design is challenging to obtain empirically by traditional methods. Optimizing its propulsion performance and launch safety is imperative. For engineering designs, the convergence process of algorithms is often passive with a high degree of randomness, and the diversity of the final results is difficult to guarantee. This paper improves the Pareto front's diversity in designing a miniaturized supersonic rocket by using a self-adaptive local diversity-preserving NSGA-II (SLDNSGA-II) in an MDO framework with an active control strategy.

Semisupervised Bacterial Heuristic Feature Selection Algorithm for High-Dimensional Classification with Missing Labels

Feature selection is a crucial method for discovering relevant features in high-dimensional data. However, most studies primarily focus on completely labeled data, ignoring the frequent occurrence of missing labels in real-world problems. To address high-dimensional and label-missing problems in data classification simultaneously, we proposed a semisupervised bacterial heuristic feature selection algorithm. To track the label-missing problem, a k-nearest neighbor semisupervised learning strategy is designed to reconstruct missing labels. In addition, the bacterial heuristic algorithm is improved using hierarchical population initialization, dynamic learning, and elite population evolution strategies to enhance the search capacity for various feature combinations. To verify the effectiveness of the proposed algorithm, three groups of comparison experiments based on eight datasets are employed, including two traditional feature selection methods, four bacterial heuristic feature selection algorithms, and two swarm-based heuristic feature selection algorithms. Experimental results demonstrate that the proposed algorithm has obvious advantages in terms of classification accuracy and selected feature numbers.

A Cross-Level Requirement Trace Link Update Model Based on Bidirectional Encoder Representations from Transformers

Cross-level requirement trace links (i.e., links between high-level requirements (HLRs) and low-level requirements (LLRs)) record the top-down decomposition process of requirements and support various development and management activities (e.g., requirement validation). Undoubtedly, updating trace links synchronously with requirement changes is critical for their constant availability. However, large-scale open-source software that is rapidly iterative and continually released has numerous requirements that are dynamic. These requirements render timely update of trace links challenging. To address these problems, in this study, a novel deep-learning-based method, deep requirement trace analyzer fusing heterogeneous features (DRAFT), was proposed for updating trace links between various levels of requirements. Considering both the semantic information of requirement text descriptions and the process features based on metadata, trace link data accumulated in the early stage are comprehensively used to train the trace link identification model. Particularly, first, we performed second-phase pre-training for the bidirectional encoder representations from transformers (BERT) language model based on the project document corpus to realize project-related knowledge transfer, which yields superior text embedding. Second, we designed 11 heuristic features based on the requirement metadata in the open-source system. Based on these features and semantic similarity between HLRs and LLRs, we designed a cross-level requirement tracing model for new requirements. The superiority of DRAFT was verified based on the requirement datasets of eight open-source projects. The average F1 and F2 scores of DRAFT were 69.3% and 76.9%, respectively, which were 16.5% and 22.3% higher than baselines. An ablation experiment proved the positive role of two key steps in trace link construction.

Tiller estimation method using deep neural networks.

This paper describes a method based on a deep neural network (DNN) for estimating the number of tillers on a plant. A tiller is a branch on a grass plant, and the number of tillers is one of the most important determinants of yield. Traditionally, the tiller number is usually counted by hand, and so an automated approach is necessary for high-throughput phenotyping. Conventional methods use heuristic features to estimate the tiller number. Based on the successful application of DNNs in the field of computer vision, the use of DNN-based features instead of heuristic features is expected to improve the estimation accuracy. However, as DNNs generally require large volumes of data for training, it is difficult to apply them to estimation problems for which large training datasets are unavailable. In this paper, we use two strategies to overcome the problem of insufficient training data: the use of a pretrained DNN model and the use of pretext tasks for learning the feature representation. We extract features using the resulting DNNs and estimate the tiller numbers through a regression technique. We conducted experiments using side-view whole plant images taken with plan backgroud. The experimental results show that the proposed methods using a pretrained model and specific pretext tasks achieve better performance than the conventional method.

Probabilistic framework for reliable optimal design of gearboxes in general-purpose industrial robots considering random use conditions

AbstractA vertically articulated robot with 6-degrees of freedom (DoF), called a general-purpose robot, can perform a myriad of different tasks within a workspace. This paper newly presents a probabilistic framework for the reliable optimal design of gearboxes used in general-purpose industrial robots, which considers random use conditions. To account for random use conditions, the start and end positions of a single motion profile are described as the random variable, which is statistically modeled as a uniform distribution based on the assumption that we have no information about the robot use pattern. Then, each sample of the random variable is converted to the corresponding motion profile by using an on-line trajectory planner. Monte Carlo simulation is implemented for the uncertainty propagation analysis, due to the heuristic feature of the on-line trajectory planner. In the design optimization formulation, the peak torque constraint and maximum bending moment constraint are described in a probabilistic manner. The system-level lifetime is calculated by combining component scale factors. The effectiveness of the proposed framework is demonstrated by examining a case study of a gearbox size problem for a 6-DoF serial industrial robot. The benefits of this study are as follows: Firstly, the proposed framework can evaluate the performance considering random use conditions. Secondly, torque reliability and bending moment reliability are newly proposed to ensure the dynamic performance of an industrial robot. Thirdly, the system-level lifetime by combining component scale factors gives more user-oriented and intuitive measure in an industrial robot design.

Heuristic and functional aspects of nomination (case-study of English anthroponymic sobriquet formations)

The article deals with functional and heuristic features of anthroponymic sobriquets in such English discourse varieties as literary discursive space, political discourse, Public Relations-discourse. Anthroponymic sobriquets represent themselves as the units of secondary nomination and linguistic manifestations of evaluative process, in which certain attitudes of linguocultural surroundings are formed and expressed. While being represented in discourse the units under analysis fulfill such basic functions as identification and evaluation, based on individual subjective characteristics that in the course of time acquire the more general character and to a certain extent tend to become possession of the linguistic cultural community. Being part of the discursive space anthroponymic sobriquets may acquire peculiar characteristics determined by communicative-pragmatic features of corresponding texts and communicative situations in question. In terms of their functional and heuristic perspectives anthroponymic nicknames serve to be the result of optimization with respect to linguistic means and intellectual effort saving, which refers to the semantic and structural peculiarities of the units under analysis, which can be regarded as both linguistic signs and signs-indexes performing discursive functions.

Diagnosis and monitoring of a person's psychophysiological state based on multidimensional transient functions of the oculo-motor system

An important scientific and practical problem has been solved, which consists in the creation of methods and tools for nonlinear dynamic identification "input-output" of the human oculomotor system (OMS). Identification is carried out based on Volterra models in the form of multidimensional transient functions (n-dimensional integrals from Volterra kernels) using innovative eye tracking technology. As test signals, visual stimuli are used, which are displayed on the monitor screen at different distances from the starting position, which formally corresponds to the action of step signals with different amplitudes at the OMS input. Experimental studies of the OMS of individuals were carried out using the Tobii Pro TX300 eye tracker. The transient functions of the first and diagonal intersections of the transient functions of the second and third order were determined based on the data of oculographic studies. The models differ from the known ones in that they provide the possibility of modeling the OMS in each interval of input signals beyond the radius of convergence of the Volterra series. The obtained multidimensional functions are used as a source of primary data in the implementation of intelligent information technology for the diagnosis and monitoring of human psychophysiological states. A set of heuristic features are proposed, which are determined using integral and differential transformations of the obtained transient functions of the OMS. The informativeness of individual signs and all their possible combinations in pairs according to the indicator of the probability of correct recognition was studied. The research results were obtained using the construction of Bayesian classifiers in different spaces of the proposed features. Figs.: 2. Refs.: 15 titles.&#x0D; Keywords: psychophysiological state, diagnosis, monitoring, oculo-motor system, identification, Volterra model, multidimensional transient functions, test visual stimuli, eye-tracking technology

Ідентифікація окуло-моторної системи людини на основі ряду Вольтерри: застосування в системі захисту інформації

The information technology of biometric identification of a person has re-ceived further development due to the use as a source of primary data of infor-mation models of the oculo-motor system (OMS) of the «input-output» type based on the Volterra series. Eye-tracking technology is used to build models. Experimental studies of the OMS of two respondents were carried out. Based on the data obtained with the Tobii Pro TX300 eye-tracker, the transient func-tions of the first, second and third orders of the OMS when applying the Volter-ra series model were determined. This makes it possible to increase the accura-cy of OMS modeling and, as a result, to increase the reliability of recognition in the space of the proposed heuristic features, which are determined using inte-gral and differential transformations of multidimensional transient functions of OMS, which greatly simplifies the identification of features and the practical implementation of the Bayesian classifier.A high variability of the transient functions of the second and third or-ders for two respondents was revealed. Thus, it seems appropriate to use multidimensional transient functions for biometric identification.A set of heuristic features are proposed, which are determined on the basis of multidimensional transient functions obtained from eye-tracking data. The informativeness of individual features and their combinations in pairswas investigated. Two-dimensional feature spaces with the maximum value of the probability of correct recognition indicator when solving the problem of biometric identification of a person were found (Pmax=0.974). The research results were obtained using the construction of Bayesian clas-sifiers in different spaces of the proposed features by means of machine learning based on the data of the formed trainingsamples

CATFPN: Adaptive Feature Pyramid With Scale-Wise Concatenation and Self-Attention

It is a typical problem in the field of object detection to simultaneously detect objects with large scale variation in one image. Recently proposed state-of-the-art object detectors generally learn pyramidal feature representation to deal with the scale variation, which has been proved effective via various feature pyramid networks. However, the majority of the feature pyramid networks based on heuristic feature fusion strategies may be suboptimal, as excess human guidance will restrict the self-learning of deep neural networks. An adaptive feature pyramid is bound to provide a significant performance boost. In this paper, we propose a novel feature pyramid network named CATFPN that consists of Scale-Wise Feature Concatenation (SWFC) module and Global Context (GC) block. The SWFC module evenly distributes semantic features for each feature layer and the GC block introduces a self-attention mechanism. As a feature pyramid network, the CATFPN can be applied to any detector based on multi-scale features. We adopt the CATFPN in typical RetinaNet and Faster R-CNN detector models, without bells and whistles, achieving 1.1% AP and 0.7% AP improvements over FPN on the MS COCO benchmark, respectively. Our competitive performance reported on the test-dev subset of COCO achieves 42.3% AP.

An optimized adaptive ensemble model with feature selection for network intrusion detection

SummaryNetwork intrusion detection system (NIDS) is a key component to identify abnormal behavior of network systems and plays an important role in preventing the occurrence of network attacks. Although a considerable number of machine learning methods have been applied in the field of intrusion detection, it is still a challenge for existing solutions to achieve a good classification performance. The existing traffic datasets generally have redundant and irrelevant features, which hinder classifiers from making more accurate predictions. Furthermore, a single classifier has limited classification performance and may not be able to achieve a better detection performance overall in the face of unbalanced multi‐category traffic data. Therefore, in order to improve the classification performance of intrusion detection models, this paper proposes an adaptive ensemble model by combining feature selection techniques and effective ensemble methods. Firstly, a heuristic feature selection algorithm (NRS‐SSA) is proposed by introducing the neighborhood dependency degree of the neighborhood rough set (NRS) into the salp swarm algorithm (SSA). Then, an improved adaptive weighted voting algorithm is designed. The SSA is introduced to optimize the weight matrix when setting the voting weight. Finally, we use the designed voting algorithm to combine the classification advantages of homogeneous classifiers and heterogeneous classifiers, respectively, and propose an M‐Tree algorithm and an adaptive ensemble model. The experimental results on multiple intrusion detection datasets show that the proposed adaptive ensemble model achieves an advanced detection level.

Noise-resistant multilabel fuzzy neighborhood rough sets for feature subset selection

Feature selection attempts to capture the more discriminative features and plays a significant role in multilabel learning. As an efficient mathematical tool to handle incomplete and uncertain information, fuzzy rough set theory has been widely used in feature selection. However, there are two pivotal issues to be addressed when adopting fuzzy rough sets for multilabel data analysis. One is how to effectively characterize the discriminative ability of the features to multilabel samples. Another one is how to alleviate the effect of data noise on feature selection. For these reasons, this study mainly studies the noise-tolerant fuzzy neighborhood rough set model for multilabel learning and its feature selection strategy. Firstly, a parameterized hybrid fuzzy similarity relation is introduced to granulate multilabel data, and the parameterized fuzzy decision is extended to multilabel learning. Then, a noise-resistant multilabel fuzzy neighborhood rough set model using inclusion relation is constructed for describing the discriminative ability of the features to multilabel samples. Furthermore, a noise-resistant heuristic feature selection algorithm titled NRFSFN is designed. Besides, motivated by local sampling and ensemble strategy, an efficient extended version of NRFSFN is designed from the perspective of data distribution, named ENFSFN. Finally, a series of comparison experiments on sixteen multilabel datasets demonstrate the effectiveness of the proposed two algorithms.

Feature selection based on double-hierarchical and multiplication-optimal fusion measurement in fuzzy neighborhood rough sets

In fuzzy neighborhood rough sets (FNRSs), uncertainty measurement performs mainly classification-hierarchical and multiplication-simple fusion, so the corresponding feature selection has advancement space. This paper aims to improve uncertainty measurement and feature selection via FNRSs. Two measurement strategies regarding class-hierarchical fusion and multiplication-optimal fusion are proposed, and three measure-based heuristic feature selection algorithms are developed. Concretely, fuzzy neighborhood self-information (FNSI) and joint entropy (FNJE) constitute two bases of heterogeneous fusion, and their multiplication fusion induces both the existing measure FNSIJE (which is based on classification-level fusion) and a new measure CFNSIJE (which is based on class-level fusion); furthermore, FNSIJE and CFNSIJE are extended to the optimal measures FNSIJEE and CFNSIJEE, respectively, by exponential parameterization. The four types of fusion measures acquire their calculation algorithms and granulation nonmonotonicity and systematically motivate four heuristic feature selection algorithms, i.e., the current FNSIJE-FS and the new CFNSIJE-FS, FNSIJEE-FS, and CFNSIJEE-FS. By using examples and experiments, relevant uncertainty measurement and granulation nonmonotonicity are validated, while the novel selection algorithms demonstrate better classification performances. This study establishes the hierarchical fusion and exponential expansion to acquire robust uncertainty measurement and optimal feature selection, and the measurement, nonmonotonicity, and selection have strong generalization for information fusion and rough-set learning.

Percussion-based quasi real-time void detection for concrete-filled steel tubular structures using dense learned features

Concrete-filled steel tubular (CFST) structures have been widely used to provide high load-bearing capacity. Percussions are expected to provide simple and effective means of detecting subsurface voids of CFST structures, caused by poor grouting compactness and harsh operating conditions. Existing percussion data processing methods consist of two steps to deal with signals with high sampling rates: heuristic feature extraction and classification. High-dimensional feature representations tend to be selected to achieve visual interpretability and performance, increasing the complexity and computational cost of the entire process. This research aims at developing a computationally efficient end-to-end framework for the percussion-based void detection of CFST structures suitable for mobile computing devices available in the field, while maintaining the performance and a certain degree of visual interpretability. The proposed approach is based on the dense learned features; First, a classifier that processes raw percussion sound signals is developed using dilated convolutions and global max-pooling. The class activation mapping (CAM) technique is then applied to convert the classifier into the dense feature extractor. Feature visualization for the experimental data shows that (i) the classifier listens to the “echo” of the percussion sound carefully, rather than the impact sound itself, and (ii) the classification results depend on both the vibration frequency and its damping ratio. Finally, the feature extractor is extended to quasi real-time void detection and classification, where the processing time duration for a single 0.1-second (50,000-point) percussion signal is 8.28 ms without leveraging advanced computing capabilities of graphics processing units, and the average testing accuracy for the fivefold cross validation is 99.81%. The results demonstrate the potential of the proposed approach for improving the efficiency of inspections.

Cardiac disease classification from ecg signals using hybrid recurrent neural network method

Medical practitioners and cardiologists frequently employ electrocardiogram (ECG) analysis to assess cardiac health. An automated ECG categorization systems have difficulty detecting and analysing the signals of numerous waves of populations, particularly electrocardiogram (ECG) data. A variety of learning strategies have been researched for the interpretation of ECG categorization. One drawback of machine learning is the use of heuristic features with shallow feature learning architectures. In order to overcome this difficulty, a deep learning approach is used to automatically learn features as opposed to using conventional handmade features. In this work, the use of Long Short-Term Memory (LSTM) and Gated Recurrent Neural Network (GRNN) techniques is proposed to develop an accurate ECG categorization and monitoring system. The purpose of the gated recurrent neural network to processing the sequential data and check the error presented or not in the ECG signals. The long short-term memory is the type of the neural network approaches to identify the complex sequence prediction problems. The GRNN and LSTM models then receive the learnt features that were first captured by the CNN model. There are no custom features required for the ECG categorization in the model. The results section lists various cutting-edge models that the CNN-LSTM and CNN-GRNN models outperformed. The CNN-LSTM and CNN-GRNN designs outperform conventional ECG classification architectures with comparable hyper-parameters, according to the comparison results.

Feature selection based on self-information and entropy measures for incomplete neighborhood decision systems

For incomplete datasets with mixed numerical and symbolic features, feature selection based on neighborhood multi-granulation rough sets (NMRS) is developing rapidly. However, its evaluation function only considers the information contained in the lower approximation of the neighborhood decision, which easily leads to the loss of some information. To solve this problem, we construct a novel NMRS-based uncertain measure for feature selection, named neighborhood multi-granulation self-information-based pessimistic neighborhood multi-granulation tolerance joint entropy (PTSIJE), which can be used to incomplete neighborhood decision systems. First, from the algebra view, four kinds of neighborhood multi-granulation self-information measures of decision variables are proposed by using the upper and lower approximations of NMRS. We discuss the related properties, and find the fourth measure-lenient neighborhood multi-granulation self-information measure (NMSI) has better classification performance. Then, inspired by the algebra and information views simultaneously, a feature selection method based on PTSIJE is proposed. Finally, the Fisher score method is used to delete uncorrelated features to reduce the computational complexity for high-dimensional gene datasets, and a heuristic feature selection algorithm is raised to improve classification performance for mixed and incomplete datasets. Experimental results on 11 datasets show that our method selects fewer features and has higher classification accuracy than related methods.