Volume 14 | Issue 5 |

 Abstract

Early problem identification is highly important as it can help guarantee effective process functioning, reduce expenses and support decision-making in areas like medicine, cyber security, banking, farming, IoT technologies and many others. The previous methods involved developing rules manually, determining thresholds and using elementary statistical methods that turned out to be insufficient to work with huge volumes of complex data. Artificial intelligence transformed this area with deep learning gaining unprecedented popularity and productivity. The computer algorithms are now able to analyse raw data without any human intervention and identify some patterns and anomalies that cannot be detected by humans. Also, the models based on deep learning can generate an extra layer of analysing numerous fragments of information allowing detecting unusual behaviour much faster than with traditional methods. It appears there are plenty of terms that should be familiar to you including CNNs, RNNs, ANNs and LSTMs, and many more models used to predict events, detect anomalies and patterns, and perform feature analysis in real life. This research paper will examine the use of artificial intelligence in identifying problems at their early stages paying attention to innovations in deep learning algorithms in particular.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

AI Based Early Detection Using Deep Learning

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

The ability to notice the initial signs of potential problems can play a key role, as it can ensure effective operation of the process itself, minimize costs and improve the decision making process within the sphere of healthcare, cyberspace, finance, agriculture, IoT devices and much more. In previous approaches the rules needed to be created manually, threshold value to be set up and basic statistical methods were utilized which proved ineffective for large amounts of complex data. The emergence of artificial intelligence revolutionized the industry bringing the idea of deep learning to the new level of popularity and efficiency. Today's computer-based algorithms are capable of analyzing vast amounts of data independently, revealing patterns and anomalies invisible to humans. Moreover, due to the deep learning models the data can be additionally analyzed with the aim of quicker detection of abnormalities. There are numerous concepts that one must be aware of such as CNN, RNN, ANN and LSTMs, not mentioning other deep learning-based models utilized to forecast events, detect problems and analyze features in real-life situations. This research paper aims to explore how artificial intelligence helps detect issues during the very early stages focusing on innovative deep learning techniques. In this paper, there are several parts that discuss neural networks that have been employed by deep learning techniques. The math processes that help train neural networks, the structure and activation function of neural networks, along with ways of improving training processes in terms of efficiency will be examined. In addition, there are some popular data sets used by researchers, software packages and their advantages and disadvantages, along with the issues related to deep learning technology today. Moreover, feature learning via deep learning can bring about better prediction results, while large data management becomes easy with artificial intelligence.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

AI Based Early Detection Using Deep Learning

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

The development of digital infrastructures has led to computer networks becoming an integral part of modern societies. Digital infrastructures serve many roles in organizations, including facilitating communication, conducting monetary transactions, storing vital data, and managing businesses. With the wide use of digital infrastructures, the probability of cyber-attacks targeting confidentiality, integrity, and availability of information systems has risen. Traditional security applications primarily use static rules and manually developed attack signatures to detect cyber-attacks. This method does not easily adapt to emerging attacks on computer networks. Hackers are always changing their tactics to avoid detection, hence rendering traditional detection methods ineffective. The growth in both volume and complexity of data in computer networks necessitates the need for automated systems to detect malicious activities. Machine learning offers a smart approach by detecting data patterns and unusual behaviour in data streams. Machine learning models undergo constant learning and enhance their ability to detect malicious activities through continuous training. This research will focus on the application of machine learning models in detecting cyber-attacks and malicious activities.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

Intelligent Cyber Attack identification using Machine Learning Techniques

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

The rapid expansion of digital technologies has significantly increased the dependency of organizations on computer networks and internet-based services. While these technologies improve efficiency and connectivity, they also introduce serious cybersecurity risks. Cyber attacks such as malware, phishing, ransomware, and denial-of-service attacks continue to evolve in complexity, making traditional rule-based security systems less effective. Conventional intrusion detection systems rely heavily on predefined signatures and manual monitoring, which limits their ability to detect new or unknown threats. Machine Learning (ML) has emerged as a promising solution to enhance cybersecurity systems by enabling automated analysis of network data and identification of abnormal behaviour patterns. ML algorithms can learn from historical data and detect suspicious activities that may indicate potential cyber attacks. This research paper explores the use of machine learning techniques for cybersecurity threat detection. It examines different machine learning algorithms, discusses data preprocessing methods, and analyses the effectiveness of ML models in detecting malicious activities within network traffic. The study highlights the advantages, challenges, and practical applications of machine learning in cybersecurity systems. The proposed approach aims to improve detection accuracy, reduce false alarms, and strengthen the overall security infrastructure of modern digital environments.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

Cybersecurity Threat Detection using Machine Learning

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

Pterygium is an ocular surface disease requiring accurate diagnosis and severity assessment for effective clinical decision-making; however, existing methods often lack detailed analysis and interpretability. This paper presents PteriGrade-Net, an explainable multimodal deep learning framework designed for automated pterygium detection and ordinal severity grading. The model integrates anterior-segment image processing, clinical features, and quantitative biomarkers. It employs advanced preprocessing followed by an Attention U-Net for lesion segmentation and biomarker extraction. These features are dynamically fused with visual representations from EfficientNet-B0 and structured clinical data using attention mechanisms to generate a unified embedding. A multi-task learning strategy optimizes three objectives: (1) binary classification (healthy vs. pterygium), (2) lesion segmentation, and (3) ordinal severity grading. Additionally, the framework enhances explainability by highlighting lesion regions and quantifying morphological characteristics, thereby improving clinical interpretability. Experimental results demonstrate superior performance compared to existing approaches in both detection and severity grading. By combining multimodal inputs with lesion-aware analysis, the proposed system aligns well with clinical workflows and offers a reliable, interpretable solution for real-world ophthalmic applications.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

Pterygium Detection; Multimodal Deep Learning; Lesion Segmentation; Ordinal Severity Grading; Explainable AI

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

Heart disease remains one of the leading causes of mortality worldwide, necessitating the development of reliable and early diagnostic systems. This paper proposes a hybrid machine learning framework designed to enhance the accuracy and robustness of heart disease risk prediction. The framework integrates multiple machines learning techniques, combining the strengths of both traditional classifiers and advanced ensemble methods to improve predictive performance. Initially, data preprocessing techniques such as normalization, missing value imputation, and feature selection are employed to ensure data quality and relevance. Subsequently, a hybrid model is constructed by integrating algorithms such as Decision Trees, Support Vector Machines, and Gradient Boosting, leveraging their complementary capabilities for improved classification. The system also incorporates feature importance analysis to identify key clinical indicators contributing to heart disease risk. Experimental evaluation on benchmark healthcare datasets demonstrates that the proposed hybrid approach outperforms individual models in terms of accuracy, precision, recall, and F1-score. The results highlight the potential of hybrid machine learning techniques in providing early, accurate, and interpretable predictions, thereby supporting clinicians in effective decision-making and preventive healthcare strategies.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

Heart Disease Prediction, Hybrid Machine Learning, Ensemble Learning, Feature Selection, Clinical Decision Support, Healthcare Analytics, Predictive Modeling, Early Diagnosis

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

Women's empowerment is essential for achieving sustainable development, especially in rural areas where socio-economic challenges restrict women's access to education, healthcare, and financial independence. The Integrated Child Development Services (ICDS), launched in 1975, has been a key initiative in tackling these issues by providing vital health, nutrition, and education services to women and children. This study explores the role of ICDS in empowering women in the Bisra Block of Sundargarh District, Odisha, by evaluating its impact on health, education, economic participation, and social awareness. Employing a mixed-methods approach, the research combines primary data from surveys, interviews, and focus group discussions with ICDS beneficiaries and stakeholders, alongside secondary data from government reports and academic studies. The findings show significant improvements in maternal and child health, higher institutional delivery rates, enhanced nutritional awareness, and increased financial independence through Self-Help Groups (SHGs). However, challenges such as inadequate infrastructure, a shortage of trained personnel, and socio-cultural barriers remain, limiting the full potential of ICDS programs. Despite these obstacles, the study emphasizes the positive correlation between ICDS interventions and women's empowerment, highlighting the need for stronger policy implementation, increased community participation, and improvements in infrastructure. By addressing these challenges, ICDS can further help reduce gender disparities and promote inclusive socio-economic development. The study concludes that a collaborative effort among government agencies, community organizations, and local stakeholders is crucial for ensuring the long-term success of women's empowerment initiatives in rural India.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

Economic Independence, Health and Nutrition, ICDS, Women Empowerment.

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

This paper presents Edumetron, a smart student productivity and academic management mobile application designed to enhance learning efficiency, track academic performance, and provide instant doubt resolution using artificial intelligence. The application integrates multiple functionalities including productivity tracking, AI-based doubt solving, analytics visualization, and gamification features such as XP and streak systems to motivate consistent learning behavior. Edumetron is developed using Flutter for cross-platform mobile development, Firebase for authentication and database management, and a Python Flask backend for AI-based query processing. The system allows students to manage their academic tasks, monitor performance metrics, and receive instant responses to their doubts through an AI-powered chatbot interface. The application is designed with a user-friendly interface and follows modern mobile application development standards. By combining productivity tools with AI capabilities, Edumetron provides a centralized and intelligent academic support system. The platform demonstrates a practical implementation of integrating mobile development with artificial intelligence to solve real-world educational challenges.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

Edumetron, Student Productivity, AI, Flutter, Firebase, Analytics, Gamification

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

Balancing authenticity and innovation in traditional interior design reflects the evolving relationship between heritage and modern creativity. Traditional elements, rich in cultural history and craftsmanship, hold symbolic and aesthetic value but must adapt to contemporary lifestyles, sustainability, and preferences. Interior designers navigate this by reinterpreting motifs, materials, and techniques within modern frameworks, preserving cultural significance while ensuring functionality. Key strategies include sustainable practices, leveraging digital technologies to enhance craftsmanship, and creating culturally resonant yet practical interiors. This ensures traditional elements remain dynamic and relevant, harmonizing cultural heritage with modern design demands and environmental considerations.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

Authenticity, Innovation, Cultural heritage, Contemporary design, Sustainability, Conservation, art restoration

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

In recent years, prolonged screen usage and sedentary lifestyles have significantly increased the prevalence of poor posture, leading to various health issues such as back pain, neck strain, and reduced productivity. This project presents a comprehensive smart posture monitoring system that integrates Internet of Things (IoT) technology with real-time computer vision and machine learning techniques to address this problem effectively.The system utilizes a personal computer equipped with a webcam to continuously monitor the user's posture. Using advanced computer vision frameworks such as OpenCV and MediaPipe, key human body landmarks are extracted and analyzed in real time. These extracted features are then processed using a machine learning model based on the K-Nearest Neighbors (KNN) algorithm, which classifies posture into categories such as "Good" and "Bad." The trained model ensures accurate and efficient posture detection under various conditions.In addition to posture analysis, the system incorporates physiological monitoring through a MAX30100 pulse oximeter sensor connected to an Arduino Nano. This sensor measures vital parameters such as heart rate and blood oxygen saturation (SpO?), providing additional insights into the user's health condition. The collected sensor data is transmitted to an ESP32 microcontroller, which acts as a communication bridge and sends the data to the cloud using the Blynk platform. The ESP32 also displays real-time data on an I2C LCD screen for local monitoring.When improper posture is detected, the system immediately triggers multiple alert mechanisms to ensure user awareness and corrective action. These include an audio alert generated using text-to-speech, an email notification with an annotated image of the detected posture, and a signal sent to the microcontroller for visual indication. The integration of these features ensures a robust and responsive feedback system.Overall, the proposed system provides a low-cost, non-invasive, and intelligent solution for continuous posture monitoring. By combining IoT, machine learning, and computer vision, it enhances user awareness and promotes healthier sitting habits. The system is highly scalable and can be further extended for applications in workplaces, educational institutions, healthcare monitoring, and smart environments.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

Index: Smart Posture Monitoring, Internet of Things (IoT), Computer Vision, OpenCV, MediaPipe, Machine Learning, K-Nearest Neighbors (KNN), ESP32, Arduino Nano, MAX30100, Blynk, Real-Time Monitoring, Posture Detection, Health Monitoring System, Webcam-Based Analysis

  License

Creative Commons Attribution 4.0 and The Open Definition