We analyse the effectiveness of differential evolution hyperparameters in large-scale search problems, i.e. those with very many variables or vector elements, using a novel objective function that is easily calculated from the vector/string itself. The objective function is simply the sum of the differences between adjacent elements. For both binary and real-valued elements whose smallest and largest values are min and max in a vector of length N, the value of the objective function ranges between 0 and (N-1) × (max-min) and can thus easily be normalised if desired. This provides for a conveniently rugged landscape. Using this we assess how effectively search varies with both the values of fixed hyperparameters for Differential Evolution and the string length. String length, population size and generations for computational iterations have been studied. Finally, a neural network is trained by systematically varying three hyper-parameters, viz population (NP), mutation factor (F) and crossover rate (CR), and two output target variables are collected (a) median and (b) maximum cost function values from 10-trial experiments. This neural system is then tested on an extended range of data points generated by varying the three parameters on a finer scale to predict both median and maximum function costs. The results obtained from the machine learning model have been validated with actual runs using Pearson’s coefficient to justify the reliability to motivate the use of machine learning techniques over grid search for hyper-parameter search for numerical optimisation algorithms. The performance has also been compared with SMAC3 and OPTUNA in addition to grid search and random search.

Typically, the image features are compared to find the similarity among the images in a content-based image clustering system. However, images with high feature similarity may be different from each other in terms of semantics. Hence, this paper proposes a novel algorithm based on unsupervised neural classifier systems for in-vivo image clustering to address the semantic gap issue. The visual features are represented using Wavelet transform and Zernike moments, and a selforganizing map is utilized for the clustering of images. The algorithm-based prototype system is trained for categorizing gastral images in the respective clusters as per the similarity. The system can be used to segment images with automatic noise reduction and rotation invariances for given images. Experiments are performed on the real gastrointestinal images obtained from a known gastroenterologist, and the results using Daubechies Wavelet Transform + Zernike Moments on LUV color scheme yield 88.3% accuracy.

In the present highly competitive telecom environment, loss of customers is a significant problem, jeopardising profitability and sustainable development. This paper investigates a data-driven methodology for churn prediction through a comparative examination of four machine learning models; FT-Transformer, TabNet, Random Forest, and XGBoost, utilising a real-world telecom dataset. The Telco Customer Churn dataset from Kaggle, initially supplied by International Business Machines (IBM) for analytical applications, is employed in this study. Adopting a comprehensive pipeline that included rigorous preprocessing, stratified cross-validation, and performance assessment through several measures, FT-Transformer has been identified as the most efficient model, with an accuracy of 84.49% and an AUC score of 0.9242. TabNet demonstrated competitive performance, surpassing conventional ensemble techniques like Random Forest and XGBoost. The findings indicate that deep learning models, especially transformer-based architectures, excel at modeling intricate patterns in customer behaviour, providing telecom companies with a robust mechanism to implement proactive retention measures.

Abstract Despite the tremendous advancements in the fields of AI in recent years, these technologies have remained understudied in the legal and forensics domain. For example, the Clearview AI system is able to match an uploaded photo with the database of billions of images collected from social media to assist the law enforcement office. This system has also assisted the Ukrainian government in April 2022 to identify living and dead, according to BBC. We propose a Photo-Synthesis system which aims to assist the criminal identification process by generating images from the available sketches. Manual identification of the criminal solely based on the sketch is cumbersome and inefficient. Photo-Synthesis system works in a two-fold manner. First, it gets a sketch input and converts it to multiple real-looking human images along with attributes-altered versions using Generative Adversarial Networks (GANs). These candidate images can directly be employed for criminal identity investigation. Secondly, the face recognition sub-system matches the facial attributes of the generated images with those present in a database of law offenders to identify the criminal. The dataset used is the publicly available Face Sketch Database provided by the Chinese University of Hong Kong. We use a series of GANs (ContextualGANs, GFPGANs, DeOldify) to implement sketch-to-image translation. The series of GAN models help to achieve a structural similarity score of 86\% between sketch and image.

Multimodal retrieval has gained much attention lately due to its effectiveness over uni-modal retrieval. For instance, visual features often under-constrain the description of an image in content-based retrieval; however, another modality, such as collateral text, can be introduced to abridge the semantic gap and make the retrieval process more efficient. This article proposes the application of cross-modal fusion and retrieval on real in vivo gastrointestinal images and linguistic cues, as the visual features alone are insufficient for image description and to assist gastroenterologists. So, a cross-modal information retrieval approach has been proposed to retrieve related images given text and vice versa while handling the heterogeneity gap issue among the modalities. The technique comprises two stages: (1) individual modality feature learning; and (2) fusion of two trained networks. In the first stage, two self-organizing maps (SOMs) are trained separately using images and texts, which are clustered in the respective SOMs based on their similarity. In the second (fusion) stage, the trained SOMs are integrated using an associative network to enable cross-modal retrieval. The underlying learning techniques of the associative network include Hebbian learning and Oja learning (Improved Hebbian learning). The introduced framework can annotate images with keywords and illustrate keywords with images, and it can also be extended to incorporate more diverse modalities. Extensive experimentation has been performed on real gastrointestinal images obtained from a known gastroenterologist that have collateral keywords with each image. The obtained results proved the efficacy of the algorithm and its significance in aiding gastroenterologists in quick and pertinent decision making.

Diabetic Retinopathy is an eye disorder that affects people suffering from diabetes. Higher sugar levels in blood leads to damage of blood vessels in eyes and may even cause blindness. Diabetic retinopathy is identified by red spots known as microanuerysms and bright yellow lesions called exudates. It has been observed that early detection of exudates and microaneurysms may save the patient’s vision and this paper proposes a simple and effective technique for diabetic retinopathy. Both publicly available and real time datasets of colored images captured by fundus camera have been used for the empirical analysis. In the proposed work, grading has been done to know the severity of diabetic retinopathy i.e. whether it is mild, moderate or severe using exudates and micro aneurysms in the fundus images. An automated approach that uses image processing, features extraction and machine learning models to predict accurately the presence of the exudates and micro aneurysms which can be used for grading has been proposed. The research is carried out in two segments; one for exudates and another for micro aneurysms. The grading via exudates is done based upon their distance from macula whereas grading via micro aneurysms is done by calculating their count. For grading using exudates, support vector machine and K-Nearest neighbor show the highest accuracy of 92.1% and for grading using micro aneurysms, decision tree shows the highest accuracy of 99.9% in prediction of severity levels of the disease.

Image classification is a challenging problem and often suffers from the bottleneck of visual features. With the ever-growing availability of multimedia data with the help of the Internet and social platforms, many images are available along with their collateral text. These linguistic keywords can be used as additional “sensors” to enhance efficiency while acting as another mode of information. This article has proposed a framework to perform the sentiment analysis on the rich textual information available from the linguistic cues of related images and incorporate them to enhance image classification. The case study has been performed on the binary classification of in-vivo gastral images and related text obtained from a known gastroenterologist. After the image classification is performed, there is a certain complex family of images that often cannot be further classified. Thus, the classification accuracy is further assisted by performing the sentiment analysis using Long Short Term Memory (LSTM) deep learning network and Bag of Words. Experimental results of the proof-of-concept have been compared with the state-of-the-art techniques to demonstrate the performance improvement of the multi-modal system.

Lately, cross-modal retrieval has attained plenty of attention due to enormous multi-modal data generation every day in the form of audio, video, image, and text. One vital requirement of cross-modal retrieval is to reduce the heterogeneity gap among various modalities so that one modality's results can be efficiently retrieved from the other. So, a novel unsupervised cross-modal retrieval framework based on associative learning has been proposed in this paper where two traditional SOMs are trained separately for images and collateral text and then they are associated together using the Hebbian learning network to facilitate the cross-modal retrieval process. Experimental outcomes on a popular Wikipedia dataset demonstrate that the presented technique outshines various existing state-of-the-art approaches.

Image segmentation is useful to extract valuable information for an efficient analysis on the region of interest. Mostly, the number of images generated from a real life situation such as streaming video, is large and not ideal for traditional segmentation with machine learning algorithms. This is due to the following factors (a) numerous image features (b) complex distribution of shapes, colors and textures (c) imbalance data ratio of underlying classes (d) movements of the camera, objects and (e) variations in luminance for site capture. So, we have proposed an efficient deep learning model for image classification and the proof-of-concept has been the case studied on gastrointestinal images for bleeding detection. The Explainable Artificial Intelligence (XAI) module has been utilised to reverse engineer the test results for the impact of features on a given test dataset. The architecture is generally applicable in other areas of image classification. The proposed method has been compared with state-of-the-art including Logistic Regression, Support Vector Machine, Artificial Neural Network and Random Forest. It has reported F1 score of 0.76 on the real world streaming dataset which is comparatively better than traditional methods.

The non-linear nature of the stock market prices and trends make it one of the most highly researched areas in the financial domain. People invest in the stock market based on multiple prediction techniques, classified into two main categories: classic methods like fundamental and technical analysis and AI-based prediction models. Both these techniques have their benefits and shortcomings. While the classical methods provide high interpretability, they may not be able to predict the complex trends of the stock market. AI-based models like random forests and neural networks can predict the trends with higher accuracy but provide little to no interpretability for their predictions, making them an uncertain tool for investment advice. In this paper, we use explainable artificial intelligence XAI to predict stock market trends and explain the predictions using two of the most prominent XAI tools, LIME and SHAP. The proof of concept and the experimental results are presented which show the promising application of machine learning in financial forecasting.

With the evolution of modern digital pathology, examining cancer cell tissues has paved the way to quantify subtle symptoms, for example, by means of image staining procedures using Eosin and Hematoxylin. Cancer tissues in the case of breast and lung cancer are quite challenging to examine by manual expert analysis of patients suffering from cancer. Merely relying on the observable characteristics by histopathologists for cell profiling may under-constrain the scale and diagnostic quality due to tedious repetition with constant concentration. Thus, automatic analysis of cancer cells has been proposed with algorithmic and soft-computing techniques to leverage speed and reliability. The paper’s novelty lies in the utility of Zernike image moments to extract complex features from cancer cell images and using simple neural networks for classification, followed by explainability on the test results using the Local Interpretable Model-Agnostic Explanations (LIME) technique and Explainable Artificial Intelligence (XAI). The general workflow of the proposed high throughput strategy involves acquiring the BreakHis public dataset, which consists of microscopic images, followed by the application of image processing and machine learning techniques. The recommended technique has been mathematically substantiated and compared with the state-of-the-art to justify the empirical basis in the pursuit of our algorithmic discovery. The proposed system is able to classify malignant and benign cancer cell images of 40× resolution with 100% recognition rate. XAI interprets and reasons the test results obtained from the machine learning model, making it reliable and transparent for analysis and parameter tuning.

The ease of access to the internet has sparked a worldwide interest in social media in recent years. This paper present a machine learning based time series analysis for post-disaster relief efforts using social media information. The situational information is gathered using Twitter from the victims and resource providers to assist government, NGOs, and off-site aid providers. For the case study, public dataset from Nepal, Italy earthquakes, COVID-19 dataset along with originally collected Twitter dataset has been considered. Vectorization has been performed using TF-IDF for machine learning model and word embeddings for the deep learning model. The performance of the extreme gradient boosting (xgboost) model is relatively superior than other techniques with 10-fold training mean accuracy of 87.17 %. Justification of the results and transparency of machine learning model is described using Explainable Artificial Intelligence (XAI) method. The experiments suggest the possibility of automation for the time series analysis for optimal relief operation management to serve the victims in the most efficient way and to control legal and administration implications.

Human beings experience life through a spectrum of modes such as vision, taste, hearing, smell, and touch. These multiple modes are integrated for information processing in our brain using a complex network of neuron connections. Likewise for artificial intelligence to mimic the human way of learning and evolve into the next generation, it should elucidate multi-modal information fusion efficiently. Modality is a channel that conveys information about an object or an event such as image, text, video, and audio. A research problem is said to be multi-modal when it incorporates information from more than a single modality. Multi-modal systems involve one mode of data to be inquired for any (same or varying) modality outcome whereas cross-modal system strictly retrieves the information from a dissimilar modality. As the input–output queries belong to diverse modal families, their coherent comparison is still an open challenge with their primitive forms and subjective definition of content similarity. Numerous techniques have been proposed by researchers to handle this issue and to reduce the semantic gap of information retrieval among different modalities. This paper focuses on a comparative analysis of various research works in the field of cross-modal information retrieval. Comparative analysis of several cross-modal representations and the results of the state-of-the-art methods when applied on benchmark datasets have also been discussed. In the end, open issues are presented to enable the researchers to a better understanding of the present scenario and to identify future research directions.

Designing an efficient fingerprint recognition technique is an ill-posed problem. Recently, many researchers have utilized machine learning techniques to improve the fingerprint recognition rate. The random forest (RF) is found to be one of the extensively utilized machine learning techniques for fingerprint recognition. Although it provides good recognition results at significant computational speed, still there is room for improvement. RF is not so-effective for high-dimensional features and also when features contain both discrete and continuous values at the same time. Therefore, in this paper, a novel similarity measure-based random forest (NRF) is proposed. The proposed technique, initially, computes both mutual information and conditional entropy. Thereafter, it uses three designed if-then rules to obtain final information measure. Additionally, to obtain feature set for fingerprint dataset, dual-tree complex wavelet transform is used to evaluate complex detail coefficients. Thereafter, ring project is considered to compute significant moments from these complex detail coefficients. Finally, information gain-based feature selection technique is used to select potential features. To prevent over-fitting, 20-fold cross validation is also used. Extensive experiments are considered to evaluate the effectiveness of the proposed technique. The comparative analyses reveal that the proposed technique outperforms the existing techniques in terms of accuracy, f-measure, sensitivity, specificity, kappa statistics and computational speed.

Medical diagnosis is a challenging procedure that involves issues such as data imbalance, insufficient labels, obscure images, redundancy and lack of effective model training directions to shrink the semantic gap between human knowledge and computer algorithms. Due to privacy norms, sometimes medical images are difficult to access and therefore, retrieval of identical existing images from an existing repository is quite useful. This paper proposes a search space to narrow down the identical images in an archive by using (i) Capsule Networks, followed by a (ii) decision fusion with Wavelet-Discrete Cosine Transform (W-DCT) and Radon Barcodes (RBC). Empirical case study has been applied on IRMA (Image Retrieval in Medical Applications) dataset, ImageCLEFMed-2009, containing 14,410 X-ray images, but the proposed method is generic, reproducible and scalable. Subjective and quantitative performance has been compared with the state-of-art and it has been found superior to yield accuracy of 92.83% and IRMA error of 124.25 for 193 class-code category. Thus, the proof-of-concept helps to improves diagnosis efficiency for automatic image retrieval and annotation by clustering similar images from the underlying repository.

Vehicles equipped with significant computing, communication and sensing (also known as “smart” vehicles), are being focused by Intelligent Transportation Systems (ITS). The primitive target of Vehicular Ad-Hoc Networks (VANETs) is to deliver safer and efficient traffic conditions by providing real time traffic conditions to automobiles and involved trusted third parties. This paper reviews eminent safety solutions to address the security aspects for VANETs. Four ingredients of this paper are (a) attacks and security mechanisms in VANETs (b) comparative analysis of security schemes based on cryptography mechanism used (c) trust management schemes based upon discrete characteristics and intrusion detection systems (d) open issues which need a thorough consideration in the future. Here we discuss how the research reflects the evolutionary growth of security attacks with its future prophesy, based upon the past developments in the area of computer security.

Deep learning has been extensively useful for its ability to mimic the human brain to make decisions. It is able to extract features automatically and train the model for classification and regression problems involved with complex images databases. This paper presents the image classification using Convolutional Neural Network (CNN) for target recognition using Synthetic-aperture Radar (SAR) database along with Explainable Artificial Intelligence (XAI) to justify the obtained results. In this work, we experimented with various CNN architectures on the MSTAR dataset, which is a special type of SAR images. Accuracy of target classification is almost 98.78% for the underlying preprocessed MSTAR database with given parameter options in CNN. XAI has been incorporated to explain the justification of test images by marking the decision boundary to reason the region of interest. Thus XAI based image classification is a robust prototype for automatic and transparent learning system while reducing the semantic gap between soft-computing and humans way of perception.

Parkinson disease is a neurodegenerative disorder of the central nerve system which affects body movements. The proposed technique selects best five machine learning models competitively, out of 25 state-of-the-art regression models to generate a robust ensemble. Data from 42 patients having early stage of Parkinson disease were collected which contains a total of 5875 voice recordings. Numerous state-of-the-art machine learning models have been explored to predict the motor Unified Parkinson’s Disease Rating Score (UPDRS) for the collected voice measures. Evaluation parameters such as correlation, R-Square, RMSE, and accuracy have been calculated for comparative analysis. Results from the ensemble model consisting of best five models have been recalculated to analyze the prediction. K -fold validation has been incorporated to measure the robustness of ensembled model. The proposed ensemble yields UPDRS with higher accuracy of 99.6% making it well suitable to assist the diagnose for Parkinson disease.