A class imbalance aware hybrid model for accurate rice variety classification

Image Classification is the task of assigning an input image to a label from a set of fixed labels. This is one of the main problems in computer vision that have many practical applications. For any classification problem, the main aim is to achieve better classification accuracy. If the classification accuracy is less, then misclassification happens and this will leads to different kinds of problems. Many of the classification models only consider the existing class instances. When a new class instance arrives the classification model not detect it properly. They actually misclassified the new class instance into an existing class instance. The proposed method therefore shows a better accurate classification and new class detection model for images. Also if needed, then the new class can be added with the model to classify correctly in the future. Recent studies show that Convolutional Neural Network(CNN) can be effectively used for image classification tasks. So here creating this better accurate classification and new class detection model based on CNN. The detection of a new class is done by looking into the trend of the softmax prediction score of class labels. In this work, the model is built for CIFAR10 image dataset. This dataset is actually a complex dataset, so creating a model for this dataset can consider as a base and extended for the classification and new class detection in other images in different applications.

In this paper we aim to investigate the trade off in selection of an accurate, robust and cost-effective classification model for binary classification problem. With empirical observation we present the evaluation of one-class and two-class classification model. We have experimented with four two-class and one-class classifier models on five UCI datasets. We have evaluated the classification models with Receiver Operating Curve (ROC), Cross validation Error and pair-wise measure Q statistics. Our finding is that in the presence of large amount of relevant training data the two-class classifiers perform better than one-class classifiers for binary classification problem. It is due to the ability of the two class classifier to use negative data samples in its decision. In scenarios when sufficient training data is not available the one-class classification model performs better.

BackgroundEffective conservation requires understanding the behavior of the targeted species. However, some species can be difficult to observe in the wild, which is why GPS collars and other telemetry devices can be used to “observe” these animals remotely. Combined with classification models, data collected by accelerometers on a collar can be used to determine an animal’s behaviors. Previous ungulate behavioral classification studies have mostly trained their models using data from captive animals, which may not be representative of the behaviors displayed by wild individuals. To fill this gap, we trained classification models, using a supervised learning approach with data collected from wild red deer (Cervus elaphus) in the Swiss National Park. While the accelerometer data collected on multiple axes served as input variables, the simultaneously observed behavior was used as the output variable. Further, we used a variety of machine learning algorithms, as well as combinations and transformations of the accelerometer data to identify those that generated the most accurate classification models. To determine which models performed most accurately, we derived a new metric which considered the imbalance between different behaviors.ResultsWe found significant differences in the models’ performances depending on which algorithm, transformation method and combination of input variables was used. Discriminant analysis generated the most accurate classification models when trained with minmax-normalized acceleration data collected on multiple axes, as well as their ratio. This model was able to accurately differentiate between the behaviors lying, feeding, standing, walking, and running and can be used in future studies analyzing the behavior of wild red deer living in Alpine environments.ConclusionWe demonstrate the possibility of using acceleration data collected from wild red deer to train behavioral classification models. At the same time, we propose a new type of metric to compare the accuracy of classification models trained with imbalanced datasets. We share our most accurate model in the hope that managers and researchers can use it to classify the behavior of wild red deer in Alpine environments.

Sensors selection for tool failure detection during machining processes: A simple accurate classification model

With the increase in the number of electronic medical records, precise classification of medical records can be use to accurately diagnose diseases. In medical texts, there are differences in the number of texts of different diseases, especially for specific diseases and a small number of samples have brought great challenges to medical text classification. In response to this problem, this paper proposes an accurate text classification model base on transfer learning combined with attention mechanism neural network, and finally obtains accurate category text through two-step classification. The model first uses the attention mechanism long-term and short-term cyclic neural network to extract the overall tumor sample from the mass of unbalanced medical record texts through the common characteristics of the tumor medical record, and uses the similarity between diseases to combine with the convolutional neural network through migration learning. The characteristics of each type of tumor disease are migrate to achieve the effect of precise classification training. The precise training model is used to finally realize the precise classification of tumor diseases. The model was tested on a private tumor medical data set, and the results showed that the accuracy of the tumor data set (precision) was increased by 5% compare to the average of the baseline optimal classification model. The increase in accuracy is important for accurate classification significance.

Introduction: Particle Swarm Optimization (PSO) was introduced in 1995 by Russell Eberhart and James Kennedy (Eberhart & Kennedy, 1995). PSO is a biologically-inspired technique based around the study of collective behaviour in decentralized and self-organized animal society systems. The systems are typically made up from a population of candidates (particles) interacting with one another within their environment (swarm) to solve a given problem. Because of its efficiency and simplicity, PSO has been successfully applied as an optimizer in many applications such as function optimization, artificial neural network training, fuzzy system control. However, despite recent research and development, there is an opportunity to find the most effective methods for parameter optimization and feature selection tasks. This chapter deals with the problem of feature (variable) and parameter optimization for neural network models, utilising a proposed Quantum–inspired PSO (QiPSO) method. In this method the features of the model are represented probabilistically as a quantum bit (qubit) vector and the model parameter values as real numbers. The principles of quantum superposition and quantum probability are used to accelerate the search for an optimal set of features, that combined through co-evolution with a set of optimised parameter values, will result in a more accurate computational neural network model. The method has been applied to the problem of feature and parameter optimization in Evolving Spiking Neural Network (ESNN) for classification. A swarm of particles is used to find the most accurate classification model for a given classification task. The QiPSO will be integrated within ESNN where features and parameters are simultaneously and more efficiently optimized. A hybrid particle structure is required for the qubit and real number data types. In addition, an improved search strategy has been introduced to find the most relevant and eliminate the irrelevant features on a synthetic dataset. The method is tested on a benchmark classification problem. The proposed method results in the design of faster and more accurate neural network classification models than the ones optimised through the use of standard evolutionary optimization algorithms. This chapter is organized as follows. Section 2 introduces PSO with quantum information principles and an improved feature search strategy used later in the developed method. Section 3 is an overview of ESNN, while Section 4 gives details of the integrated structure and the experimental results. Finally, Section 5 concludes this chapter.

Factors controlling the distribution of shelf sand as a resource, a component of reef ecosystems, and a dynamic substrate are poorly understood. An initial step in understanding sand accumulation in each of these roles is to identify its areal extent and change through time. Digitized aerial photographs and digital images provide common, inexpensive data sets that are generally underutilized for the purpose of marine substrate classification. Here we use only two bands, blue and green (470 and 550 nm), to demonstrate the utility of simple aerial photography in classifying marine substrate. Although these two are acquired from a hyperspectral data set, they represent blue and green in an RGB image such as commonly available in digitized aerial photographs. We add as a third band the second eigenchannel of a principal components analysis of these bands. Using an artificial neural network classification model, we identify submarine and subaerial sandy substrate in a digital image of a detached reef island in the Red Sea, Gezirat Siyul, Egypt. With careful selection of training and test groups, using small percentages of the total classified image, we create an efficient and accurate classification model. The model, trained to identify two classes, “sand” and “other than sand,” produces a classified image that provides sand locations and approximate areal coverage. Confusion matrices for both training and testing groups have user's accuracies in the 90 percentiles, indicating accurate pixel classification.

Tumor-related epilepsy (TRE) refers to the condition in which primary brain tumors cause recurring seizures. A model that classifies brain tumors as epileptogenic or non-epileptogenic could improve prognosis and treatment methods for TRE. This study aims to identify which MRI sequences and machine learning algorithms (MLAs) could be used to build the most accurate epileptogenic tumor classification model. T1W, T2W, T2W FLAIR and T1W contrast-enhanced scans were acquired from 24 glioma patients, 8 with and 16 without pre-operative epilepsy. A total of 88 features were extracted from the MRI sequences, including tumor location, volume, and several first order textural features derived from gray level co-occurrence matrices (GLCM). Each feature was then used as a predicting variable for 9 MLAs (7 variants of support vector machines (SVMs) and 2 variants of logistic regression) to construct classification models. The top 11 classification models had testing accuracies above or equal to 75%. These models all used SVM variants instead of logistic regression variants. The classification model that used tumor location, and the one that used tumor volume, had testing accuracies of 100% and 87.5% respectively. The 9 other top classification models used GLCM features extracted from the contrast T1W sequence.Clinical Relevance—Our study showed that models which used SVMs were more accurate at classifying tumors by epileptogenicity than those that used logistic regression variants, and contrast T1W radiographic features could also be used in epileptogenic tumor classification models.

The key to successful early recovery and treatment of breast cancer in today's healthcare system is an accurate and prompt diagnosis. Over the last several years, the IoT has undergone a transition that makes it possible to analyse both real-time techniques. Medical diagnostics are aided by the Internet of Medical Things, which connects various medical equipment and artificial intelligence applications with the healthcare network. Most women with breast cancer don't make it because the disease isn't detected early enough using today's best methods. Therefore, doctors and scientists are confronted with a significant challenge in recognizing breast cancer at an primary stage. We present a medical IoT-based diagnostic system that can distinguish between patients with cancer and those without it in an Internet of Things setting. Malignant vs benign categorization is performed using an unique transfer learning technique called BERT, which is based on a previously learned language model. In particular, this research looks at how well novel fine-tuning approaches based on transfer learning might improve BERT's capacity to capture significant context. This research improves the BERT model's classification accuracy by using a Black Widow-meta-heuristic Optimization (NBW-MHO) feature selection strategy to refine feature selection from the breast cancer dataset. The WDBC dataset served as a testbed for the suggested method. The suggested model's classification accuracy using the BERT model and NBW-MHO was 95.20 percent.

Peatland in tropical region is a major CO2 emission source because of peat decomposition and forest fire by human induced activities. Remote sensing is effective tool to monitor environmental condition of peatland and forest ecosystem in peatland. A pixel-based approach is one of the most attractive choices for forest type classification or biomass prediction. The traditional method, however, is not sufficient for using spatial information. The spatial information, such as image texture, is an important factor for identifying objects or types, because a pixel is not independent of its neighbors and its dependence can be useful for classification and biomass prediction in forest regions. In this paper, we used combined data of spectral and spatial information from hyperspectral data (Hymap) to develop a more accurate classification or biomass prediction model. The spatial information was texture data by using Grey Level Co-occurrence Matrix (GLCM) texture measures. Sparse discrimination analysis (SDA) was applied for the classification model, and LASSO regression was applied for the biomass prediction model. The results were compared to find out how the spatial information enhances the classification and biomass prediction. According to the accuracy assessment, both classification and biomass prediction model derived from the combined data performed high accuracy.

Ovarian cancer is one of the leading causes of death worldwide among the female population. Early diagnosis is crucial for patient treatment. In this work, our main objective is to accurately detect and classify ovarian cancer. To achieve this, two datasets are considered: CT scan images of patients with cancer and those without, and biomarker (clinical parameters) data from all patients. We propose an ensemble deep neural network model and an ensemble machine learning model for the automatic binary classification of ovarian CT scan images and biomarker data. The proposed model incorporates four convolutional neural network models: VGG16, ResNet 152, Inception V3, and DenseNet 101, with transformers applied for feature extraction. These extracted features are fed into our proposed ensemble multi-layer perceptron model for classification. Preprocessing and CNN tuning techniques such as hyperparameter optimization, data augmentation, and fine-tuning are utilized during model training. Our ensemble model outperforms single classifiers and machine learning algorithms, achieving a mean accuracy of 98.96%, a precision of 97.44%, and an F1-score of 98.7%. We compared these results with those obtained using features extracted by the UNet model, followed by classification with our ensemble model. The transformer demonstrated superior performance in feature extraction over the UNet, with a mean Dice score and mean Jaccard score of 0.98 and 0.97, respectively, and standard deviations of 0.04 and 0.06 for benign tumors and 0.99 and 0.98 with standard deviations of 0.01 for malignant tumors. For the biomarker data, the combination of five machine learning models-KNN, logistic regression, SVM, decision tree, and random forest-resulted in an improved accuracy of 92.8% compared to single classifiers.

MLapRVFL: Protein sequence prediction based on Multi-Laplacian Regularized Random Vector Functional Link

Lung Cancer is one of the hazardous diseases that have to be detected in earlier stages for providing better treatment and clinical support to patients. For lung cancer diagnosis, the computed tomography (CT) scan images are to be processed with image processing techniques and effective classification process is required for appropriate cancer diagnosis. In present scenario of medical data processing, the cancer detection process is very time consuming and exactitude. For that, this paper develops an improved model for lung cancer segmentation and classification using genetic algorithm. In the model, the input CT images are pre-processed with the filters called adaptive median filter and average filter. The filtered images are enhanced with histogram equalization and the ROI(Regions of Interest) cancer tissues are segmented using Guaranteed Convergence Particle Swarm Optimization technique. For classification of images, Probabilistic Neural Networks (PNN) based classification is used. The experimentation is carried out by simulating the model in MATLAB, with the input CT lung images LIDC-IDRI (Lung Image Database Consortium-Image Database Resource Initiative) benchmark Dataset. The results ensure that the proposed model outperforms existing methods with accurate classification results with minimal processing time.

A new model is proposed to automatically classify sleep stages using heart rate variability (HRV). The generative model, based on the characteristics that the distribution and the transition probabilities of sleep stages depend on the elapsed time from the beginning of sleep, infers the sleep stage with a Gibbs sampler. Experiments were conducted using a public data set consisting of 45 healthy subjects and the model's classification accuracy was evaluated for three sleep stages: wake state, rapid eye movement (REM) sleep, and non-REM sleep. Experimental results demonstrated that the model provides more accurate sleep stage classification than conventional (naive Bayes and Support Vector Machine) models that do not take the above characteristics into account. Our study contributes to improve the quality of sleep monitoring in the daily life using easy-to-wear HRV sensors.

LitefusionNet: Boosting the performance for medical image classification with an intelligent and lightweight feature fusion network