Computational Intelligence in Communications and Business Analytics

Alzheimer’s disease (AD) is an incurable neurodegenerative disease which is one of the leading causes of death in elderly people. Early and accurate detection of AD is vital for appropriate treatment. AI-based automated techniques are widely used to help early diagnosis of AD. In recent years, machine learning and deep learning has become the preferred method of analyzing medical images, and it has also attracted a high degree of attention in AD detection. Researchers have proposed many novel approaches for automated detection and gradation of the disease. The success of any such approach depends on the appropriate selection of pre-processing, biomarkers, feature extraction, and model architecture. This paper presents a review of the efficacy of different methods used by the researchers for these components with the aim to understand the state-of-the-art architecture. A comparative analysis of their advantages, disadvantages, and performance accuracy is reported.

Landslides are one such spontaneous natural disaster which has the potential to effect human lives as well as economic property in any region. The continuous exploration of qualitative and quantitative approaches augmented with RS-GIS methodologies for assessing landslides have given new dimensions to the area of research. The main obstacle of these methodologies remains in the fact that the hypothesis needs to be taken as true even before the analysis. To overcome the same, with the development of machine learning methodologies, the exhibits are found to be more objective in terms of quantification as well as the quality of research. The current assessment was aimed to evaluate the execution of two machine learning approaches namely; logistic regression and random forest for the assessment of slope instability in Darjeeling Himalayas. Eight geo-environmental factors were taken into consideration for the said assessment. The susceptibility models prepared with the said approaches were substantiated using receiver operating characteristics curves. The assessed accuracies were 76.8% and 79.7% for logistic regression and random forest respectively with the dataset(training and validation), the accuracies measured were 76.8% and 77.7% respectively.

Identification of text parts and non-text parts present in offline unconstrained handwritten manuscripts is an essential step toward the construction of an effective optical character recognition (OCR) system. To address the said issue researchers mostly extracted handcrafted features which capture the texture information in order to recognize text or non-text components separately. In presence of noise, these types of feature descriptors badly suffer. Therefore, in this paper, a Convolutional Neural Network (CNN) is designed to separate these extracted components. To evaluate the developed model, an in-house dataset of 150 pages is created. In this dataset, the present model has achieved 85.07% accuracy. The performance of the present model is compared with three recent works where it has outperformed these existing works.

The three-frame difference is a renowned tactic for detecting moving items. According to the idea, the existence of a moving object can be inferred by removing three subsequent image frames that display the moving object’s edges. However, information is lost as a result of the method because these edges do not convey all of the information about the moving item. To obtain all of the information about a moving object, post-processing techniques like morphological operations, optical flow, and combining these techniques must be used. In this study, we introduce a novel method to detect moving objects in video sequences without any post-processing steps, dubbed Selected Three Frame Difference (STFD). We first provide an algorithm that selects three images while accounting for the local maximum value of frame disparities rather than employing successive. The logical operator is applied to three frames of three different picture variations that contain non-overlapping object frames. We mathematically show that the entire moving object is always discernible in the second image that was selected. We investigated the proposed strategy on a dataset collected in our lab and a public benchmark dataset. We compared the effectiveness of our approach to the three-frame difference method and background subtraction-based traditional moving object recognition methods on a few sample videos selected from different datasets.

Diabetic retinopathy (DR) is an ophthalmological ailment wherein the diabetic individuals suffer from the formation of blockages, lesions, or hemorrhages predominantly in light-sensitive portion of said retina. Because of the increase in blood sugar, vascular blockage drives new vessel creation, giving rise to mesh-like patterns. As lack of timely treatment of DR results in vision loss, early diagnosis and professional assistance plays a crucial role. This can be achieved with a computer-aided diagnostic (CAD) system based on retinal fundus images. Various steps are involved in a CAD system, including as the detection, segmentation, and categorization of abnormalities in fundus images. This study is an effort to expedite the first screening of DR so as to meet the need of the increasing population of diabetic patients in the future. On publicly accessible datasets, we have trained and validated reliable classification algorithms enabling timely detection of DR. Convolutional neural networks (CNN)-based advanced deep learning models are used to fully use data-driven machine learning techniques for this purpose. We also defined the issue as the detection of DR of any grade (Grades 1–4) vs. No-DR in a binary classification (Grade 0). For training the models, we used 56,839 fundus pictures from the EyePACS dataset. On a test set from EyePACS, the models were put to the test (14,210 images). As compared to the established methods, experimental findings demonstrate superior outcomes through DenseNet with pre-trained weights. In the model’s evaluation on the EyePACS datasets, it achieved good results with an of 97.55% in binary and 78% in multiclass-classification.

A radial network is one that traverses a network without connecting to another source of supply. It is utilised for remote loads, such as in rural areas. For the load flow analysis of radial distribution systems, various forward-backward sweep techniques exist. This study explains a novel approach to load flow analysis for radial distribution systems. Encouraged by whales’ use of bubble-net hunting, WOA imitates humpback. The suggested technique is applied on IEEE 33-bus and IEEE 69-bus balanced radial distribution test networks to validate performance in tackling the described problem. The results show that the suggested approach produces workable and efficient solutions and may be successfully substituted for in real-world power systems for radial network load flow analysis. Additionally, to the best of the authors’ knowledge, this is the first report on the use of WOA in resolving the optimal DG.

Hundreds of continuous bands make up a hyperspectral image. All the bands are not equal important. Some of the bands are significant and others are redundant. Band reduction is a typical step before further processing. Instead of attempting to handle the complete information set without losing crucial data, it is essential to select the most valuable bands. Using traditional band selection techniques, the predetermined number of dimensions are selected from the hyperspectral image. In this article, we propose a novel single-layer neural network and a genetic evolutionary approach to reduce a hyperspectral image’s high dimension. The process involves selecting the two bands with the lowest correlation in each iteration and eliminating two redundant bands. The suggested framework eliminates the unnecessary bands from a hyperspectral image and then chooses the ideal number of the most crucial bands.

This article presents an elementary and efficient nature inspired optimization technique namely African vulture optimization algorithm (AVOA) to solve economic load dispatch problems. To make it more cost-effective, wind turbines have been incorporated with the existing thermal generating plants. The stochastic behaviour of wind is considered here. AVOA has been implemented for single objective fuel cost minimization. For demonstrating suitability and scalability of this proposed approach in case of large-scale and real world scenerio, it has been tested against IEEE 6-bus, IEEE 40-bus and IEEE 140-bus network and the outcomes are analyzed against results found by other heuristic approaches that were being used recently. The results clearly shows the potential AVOA has in achieving optimal solution.

This paper presents a wireless battery charging scheme for electric vehicle. Here, electrical source can be either conventional power supply or solar photovoltaic power source. During sufficient solar radiation level, charging power is wirelessly fed from solar PV module through the controlling of converter under maximum power point tracking (MPPT) mode. Otherwise, battery is charged from conventional power source via wireless method. Under passing cloudy condition or presence of nearby objects like buildings or trees, partially shaded PV module fails to supply maximum electrical power for charging battery bank. Here, the conventional MPPT algorithm is unable to track the global peak power under presence of multiple peaks on power-voltage characteristic of partially shaded PV module. Thereby, this reduces power conversion efficiency of PV module under partial shading condition. In the context, Grey Wolf Optimization (GWO) based MPPT method is proposed to maximize charging power of battery. This facilitates efficient and rapid charging action of battery bank in EV application.

Food scarcity will be a threatening problem in front of the global civilization due to huge growth in world population and reduce in world agricultural land covers. Agriculture depends on several factors like climate, soil conditions, irrigation, fertilization, condition of pests. The increase in carbon footprint due to civilization adversely affects the worldwide climate which causes unexpected floods, droughts and increase in pests directly affects the productivity and quality of agricultural products. We can increase the productivity of agricultural sector by analyzing and predicting the data of external parameters like carbon footprint, rainfall information, moisture information, soil information by predicting flood, drought, pest movement and other factors. In this article, we tried to perform the prediction of rainfall and carbon-footprint and used regression analysis for finding the correlation between Indian agricultural data containing carbon footprint and rainfall over Indian geography which can helps to increase the indian agricultural product.

The chest X-ray (CXR) images of healthy patients and patients with COVID-19 are clustered into distinct classes using the mean structural similarity index measure (SSIM). SSIM is intrinsically similar to the human visual system (HVS) and has potential for extracting the information for structural changes in the image to perceive the distortions. The proposed approach is based on local statistical parameters like mean, variance etc., to extract structural information through SSIM. This information is subsequently used for CXR image differentiation, akin to the clinician's visual inspection of these images. As a feature extractor, SSIM is found to effectively classify and characterize COVID-19 patients from the healthy ones from analysis of the CXR images. Our approach of classifying CXR images, based on a single comparative parameter with the use of an ensemble tree classifier, leads to an accuracy equivalent to the recently developed methods using a variety of convolutional neural network (CNN) approaches and is computationally faster. We obtained an accuracy of 97.7% for our proposed models. The obtained results are corroborated through the statistically reliable analysis from the receiver operating characteristic (ROC) curve and confusion matrix. The comparative SSIM index enables the effective use of larger data points for the classifier’s robust training due to cross-correlation between healthy subjects and diseased ones, yielding higher classification accuracy. Our proposed method may find clinical application for classifying patients of COVID-19 using CXR images.

This paper has the sole purpose of showing the fact that religious sentiment detection holds an important place in the industry and our efforts have been totally concerned with easing the problems of the industry which have been existing to date. Our research has been focused on the shortcomings of various previous methods which have been suggested by previous researchers to classify religious sentiments. There has never been any single application that can classify the sentiments present in a given block of religious text by analyzing only the religious text. We have designed the model in such a way that the users will not have to specify the religious texts and filter them out. The application will reject all the nonreligious texts and provide the desired outcome after analyzing the given religious text which is provided by the user. The model works on the basis of Natural Language Processing and it is able to handle a large amount of data. It is trained using the data sets of the 12 main religions of the world and it is able to perform predictive analysis of the input text since the model is trained using RNN and LSTM algorithms. We have also used the KNN algorithm in the testing phases of the model. A brief analysis of the time complexity along with the comparison of performance evaluation among the different methods have also been discussed in this paper. In the results that we received, it can be clearly seen that we have achieved a minimum loss of 0.083, and the highest accuracy value of the model is found to be 99.8%, This study evaluates the different approaches that can be used to perform sentiment analysis on religious texts and provides a landmark for future researchers to continue improvements in this field. Our research paves the way for future researchers to work more on the untouched portions of sentiment analysis and its applications in real life. In this way, these extensive technologies can be put to better use. We believe that our work and our results might be able to help the people in common to get rid of the harmful and malicious effects of certain religious texts and they would be able to recognize the religious texts which carry good value or provide comfort to them.

The advancement of modern technology has enabled the diagnosis of various skin diseases through image processing. Researchers face significant challenges when it comes to utilize image processing tools for skin disease analysis. One particularly serious disease is Melanoma, a type of cancer originating from melanocytes. The primary objective of this research article is to develop a machine learning classification-based algorithm for melanoma detection.

The complexities associated with analyzing melanoma skin disease can be mitigated by employing an effective classification technique. In this proposed model, two machine learning (ML) classification algorithms, namely the Probabilistic Neural Network (PNN) and Support Vector Machine (SVM), are utilized for disease detection. These algorithms are employed to differentiate between melanoma-affected skin and normal skin.

To establish the machine learning algorithms, feature selection is performed using Factor Analysis (FA), and the dermenetNZ.org dataset is utilized. The performance of the two classification algorithms is compared using standard performance metrics. Through comparative analysis, it is demonstrated that the Probabilistic Neural Network classifier outperforms the Support Vector Machine in the classification of melanoma skin disease. Overall, this research article showcases the efficacy of a machine learning-based approach for melanoma detection, with the Probabilistic Neural Network exhibiting superior results compared to the Support Vector Machine classifier.

Satellite Clouds have a significant role in the weather system and climate change, and the distribution of clouds is always strongly tied to a particular meteorological phenomenon. In this paper, an automatic identification of cloud types is proposed using a hybrid approach of convolution neural network (CNN) and bidirectional character based long short-term memory (LSTM). The large-scale cloud image database for meteorological research (LSCIDMR) of the ground truth images related to weather types is used as the input for the proposed work. Three types of CNN models, such as inception v3 network, Vgg-16 and Alexnet, are used separately and subsequently, the results are compared, in terms of precision, recall, and F1 score, to obtain the best among them. The LSTM is trained with our self-trained dictionary having tokens. The image features and single character are merged into a single step. It produces the output as the next character to come and so on.

Tracking market fear in distress periods is a highly challenging and essential task of paramount practical relevance. If the future figures of market fear can be predicted in conjunction with explaining the dependence structure on predictor variables, market players at different levels can be benefited. The current work endeavors to model the Chicago Board Options Exchange’s Volatility Index (CBOE VIX) of the US, reflecting the extent of market fear in the futures market through the lens of applied predictive modeling and explainable artificial intelligence (AI). The methodological framework deploys two advanced forecasting tools, namely, Facebook Prophet and Uber Orbit, to gauge the temporal pattern of the CBOE VIX. The exercises have been carried out across different regimes characterized by varying degrees of volatility and uncertainty. It is revealed that the market fear in the US was relatively more predictable during the Pre-COVID-19 phase. The outcome of explainable AI analysis using Shapley additive explanations (SHAP) and accumulated local effect (ALE) plots indicates the past information of CBOE VIX exerts significant predictive influence, which largely explains the variation.

The use of AI is becoming increasingly widespread in medical diagnosis. Recently, many decision-making systems have used the Artificial Neural Networks (ANN) model to train the ANN’s weight and biases to get the lowest error function and highest accuracy. In this concern meta-heuristic based optimization technique play an important role. Already various optimization techniques have been applied to train an ANN’s weight and bias. But due to improper balancing between exploration and exploitation they fail to give the global optima. To overcome this issues, this study used a new stochastic-based optimization algorithm the Sine Cosine Algorithm (SCA). The mathematical formulation of SCA is based on trigonometric functions, sine and cosine. However, sometimes slow convergence is the main disadvantage of the basic SCA algorithm. This paper proposes a modified SCA optimization technique called Chaotic SCA(CSCA) to train the control parameters like weights and biases of a single-layer ANN by integrating chaotic into SCA to expedite the convergence speed. The performance of the above algorithm is examined and verified using The Pima Indian data set. The experiment revealed the outperformance of CSCA than the other algorithms.

One of the most common oncological types is oral cancer. Although medical technology has advanced at a phenomenal rate, high fatality has been observed in developing countries due to the lack of early stage diagnosis of the disease. The most fundamental symptom being the prolonged inflammation in the mouth areas. Cancers in the tongue, lips, cheeks, the floor of the mouth, hard and soft palates, sinuses, and pharynx (throat) are all considered oral cancers. This study focuses on early stage diagnosis of the disease using deep learning frameworks. It will offer a more thorough understanding of the disease and help experts make judgments about diagnostic and treatment options that are well-versed. We have used a deep learning framework based on the modified Convolutional Neural Network (CNN) that uses different sizes of hidden layers. The dataset comprised histopathology images. Histopathology datasets have the potential to transform the field of medical research. By feeding a histopathology dataset into a deep-learning framework, researchers can rapidly and precisely classify patterns in the data that would otherwise be difficult or impossible to detect. It could lead to faster diagnosis of diseases and more effective treatments. A total of 8000 images (4000 for each category of the cancers) are used for result analysis. Per epoch, the testing loss likewise diminishes gradually. As a final result, at 30 epochs, it has reached the highest accuracy of 97.6%. The convolutional neural network exhibits result which fare better than peer proposals in literature.

Antinuclear antibody patterns are used as an important screening technique to diagnose autoimmune disorders. The rising prevalence of autoimmune conditions, such as connective tissue diseases, has resulted in an increase in the production of antinuclear antibodies (ANA). Unavailability of expert pathologist delay the analysis and interpretation of ANA patterns in many places. Automated analysis of ANA pattern can reduce the time of pathological investigation and help doctors to plan for the treatment. This work proposes a convolutional neural network based model that can classify the ANA pattern into four different categories - mitotic, nuclear, cytoplasmic, and negative. The model trained with relatively fewer number of samples has performed satisfactorily while being trained and tested with ANA dataset. It exhibits a relatively good F1 score of 0.97.

We propose a method by integrating image visibility graph and deep learning (DL) for classifying COVID-19 patients from their chest X-ray images. The computed assortative coefficient from each image horizonal visibility graph (IHVG) is utilized as a physical parameter feature extractor to improve the accuracy of our image classifier based on Resnet34 convolutional neural network (CNN). We choose Resnet34 CNN model for training the pre-processed chest X-ray images of COVID-19 and healthy individuals. Independently, the preprocessed X-ray images are passed through a 2D Haar wavelet filter that decomposes the image up to 3 labels and returns the approximation coefficients of the image which is used to obtain the horizontal visibility graph for each X-ray image of both healthy and COVID-19 cases. The corresponding assortative coefficients are computed for each IHVG and was subsequently used in random forest classifier whose output is integrated with Resnet34 output in a multi-layer perceptron to obtain the final improved prediction accuracy. We employed a multilayer perceptron to integrate the feature predictor from image visibility graph with Resnet34 to obtain the final image classification result for our proposed method. Our analysis employed much larger chest X-ray image dataset compared to previous used work. It is demonstrated that compared to Resnet34 alone our integrative method shows negligible false negative conditions along with improved accuracy in the classification of COVID-19 patients. Use of visibility graph in this model enhances its ability to extract various qualitative and quantitative complex network features for each image and enables the possibility of building disease network model from COVID-19 images which is mostly unexplored. Our proposed method is found to be very effective and accurate in disease classification from images and is computationally faster as compared to the use of multimode CNN deep learning models, reported in recent research works.

Today Potato becomes most well-known crops in world. Now Plant crop disease detection has transferred as an operative research domain. As per enhancement of requirements of methods and demands for detection of diseases of crops are crucial part of agriculture. Many disease affects the perfect enhancement of plants of potatoes. Some Observable problems are very much visible in potato plants leaf areas of affected regions As Early (EB) and Late (LB) Blight. Particularly, image based approach offers the way of gathering knowledge regarding plants for quantitative analysis. In case of other side, manual detection of crop diseases needs more work effort, expert domain persons, execution time higher. Therefore, integration of image processing and machine learning is required to enable the diagnosis of leaf images with disease. CNN is used for image Detection and Analysis of potato diseases and gives the best result than other classifier. Here some classifiers are used for this research paper such as SVM, Random Forest, Logistic Regression & Sequential model. In this proposed work, the model validation, training is done using CNN to identifying and extraction of necessaryinformation of used datasets and for determining that leaf are affected or not. This model achieved accuracy of 97.92% that indicates the suitable outcomes for identifying the crop diseases.

Heart conditions are classified as diverse illnesses with numerous subgroups. To make patient clinical treatment easier, early heart disease diagnosis and prognosis are crucial. Although much research has been done to predict the cause of heart disease. We have tried to build a heart disease detection system using machine learning algorithms. The basic goal of work is to detect whether a person is suffering from heart disease or not. We have also built a GUI using HTML and CSS for our front end and integrated both using flask for the back end. The user interface is available at https://​github.​com/​SayanKumarBose/​FinalYearProject​.​git.

Millions of people play and enjoy the game of cricket, however, classifying the diverse batting style and postures used frequently by the batsman during a cricket match has always been a difficult proposition. Owing to the immense overlap between postures and various styles of the same shot, it gets extremely harder. Sports experts, trainers, and coaches must learn more about the variety of approaches used by each batsman in both international and local matches in order to guide the team in its entirety and ensure that the training program of cricket players is planned and executed to its fullest potential. The work in this paper thrives on a hybrid deep learning approach that combines convolutional and recurrent neural networks for classifying ten (10) types of cricket shots from match videos. To establish a baseline, a sports CrickShot10 [1] dataset and an open-source cricket video dataset are used. Automatic feature extraction is handled by a hybridized form of convolutional neural network (CNN [11])- recurrent neural network (RNN) combined. Long temporal dependencies are handled by a Gated Recurrent Unit (GRU [12]). It is further improved by adding a Self-attention [20] module that is introduced to the hybrid module to facilitate a semi-supervised approach to extract the key frames from the video. This idea is intended to address the architecture’s inconsistent behaviour while processing somewhat long videos, and their inability to give “correct/relevant” frames priority. When results are compared to other modules, they show good accuracy values. Here we focused on ‘Accuracy’ instead of other evaluation metrics as this is a task of simple classification.

Image quality degrades due to various reasons. In some circumstances, different weather conditions like fog, mist or rain have an impact on image visibility. Dust and pollution in the air can reduce the clarity of images taken outside. Thus, input images with poor visibility may reduce the effectiveness of computer vision related applications. The automated traffic monitoring systems, computer vision based smart systems used in different vehicles are few examples of such applications. Image restoration is essential of such applications for accurate implementation. In the proposed work, an end-to-end network is designed to restore images affected by rain, fog, dust and pollution. An integrated Convolutional Neural Network (CNN) with channel attention method is proposed for image restoration. In the proposed work a CNN is designed to reduce the loss between input degraded image and clear ground-truth image. Channel attention technique based on Style-based Recalibration Module (SRM) is applied on convolutional feature maps to improve the visibility of the restored image. The model is trained on a synthesized dataset and it is then evaluated on both synthesized and real-world out-door and traffic images. The experimental results demonstrate that the proposed method is more effective to several state-of-the-art methods both quantitatively and visually.

Skin diseases are generally normal around the globe, as people get skin diseases because of inheritance and natural elements. Dermatologists rely heavily on visual examinations to diagnose skin diseases. However, this method can be inaccurate and time-consuming. The development of the technique in deep learning (DL) and the availability of GPUs can expand and improve the quality of computer-aided disease diagnostics systems. This paper proposes a DL-based smartphone application to aid a dermatologist in diagnosing skin disease in real-time. In the proposed work, we first used fine-tuned DenseNet201 for feature extraction and SVM to accurately classify normal and abnormal skin. A Deep Ensemble CNN (DECNN) framework with DenseNet-201, Resnet50, and MobileNet is further used for skin disease categorization if abnormal skin is detected. The experimental outcome of this study demonstrates that the framework can achieve an 84% accuracy in skin disease classification and outperform existing state-of-the-art works. The simple implementation and acceptable accuracy of the proposed method can be helpful for dermatologists in the diagnosis of skin disease.