Advances in Data-Driven Computing and Intelligent Systems

Inhaltsverzeichnis

1. Frontmatter

2. A Novel Outlier Detection Approach Using ECOD, LUNAR and Logistic Regression

   Purav Patel, Nalini S. Jagtap

   Abstract

   A significant number of methods have been devised for anomaly detection. Traditional methods like LOF, IForest, CBLOF, etc. have a strong performance and simple methodologies, making them very popular. However, the recently proposed methods of LUNAR and ECOD trump them with their better performance, where ECOD is a simpler, faster, and more interpretable method with a profound ability to detect global anomalies, and on the other hand, LUNAR introduces the trainability of parameters as a tremendous advantage and unifies the local outlier detection methods. These methods allow for the introduction of a new ensemble method that utilizes LUNAR and ECOD as base models, combined with Logistic regression as a meta model. This study introduces a technique that surpasses the original foundational techniques and other widely used outlier detection methods. The results showed that the proposed method had an 8.94% increase in accuracy and a 10.41% improvement in AUC score compared to the base models of ECOD and LUNAR. It consistently produces reliable and improved results compared to other algorithms.

3. Automating the Billing Process for Medicines Using Optical Character Recognition and Named Entity Recognition

   Dhruv Patel, M. Dominic Savio, V. Pandiyaraju, P. Anandan

   Abstract

   Billing for medicines in healthcare environments is often a manual and error-prone process, especially under high pressure. This research proposes an automated system leveraging Named Entity Recognition (NER) and Optical Character Recognition (OCR) using the Spacy library to address these challenges. The approach involves extracting key details from images of medicine labels through OCR, followed by classifying these details into entities like batch numbers using NER. The identified information is then matched against a pre-populated CSV database to retrieve drug names, prices, and expiration dates. This data is used to automatically generate invoices, significantly reducing manual intervention. The system is designed to handle poorly formatted text and multilingual inputs, making it versatile in diverse healthcare settings. The integration of OCR and NER not only boosts billing efficiency but also improves accuracy compared to traditional methods, paving the way for more advanced automated billing solutions in healthcare.

4. Prioritizing Risks in IoT-Enabled EdTech Platforms: A Fuzzy AHP Approach to Maximize User Satisfaction

   Ankita Priti Roy, Kerena Anand, D. Halaswamy, N. Elangovan

   Abstract

   The integration of the Internet of Things (IoT) in educational technology (EdTech) platforms offers personalized, adaptive learning but introduces significant risks. This study identifies and prioritizes these risks using the Analytic Hierarchy Process (AHP) and fuzzy AHP techniques. Seven industry experts provided input, complemented by a comprehensive literature review. The analysis reveals ethical considerations (30.17%) and data privacy/security (29.22%) as top concerns, followed by regulatory compliance (12.91%), high implementation costs (10.99%), and technical expertise requirements (8.37%). Surprisingly, scalability concerns (1.70%) and data accuracy/reliability (2.63%) rank lower. These findings emphasize the need for a human-centric approach in IoT-enhanced EdTech deployment, focusing on responsible implementation and regulatory adherence. The study provides valuable insights for EdTech companies and educational institutions, guiding strategic decision-making to enhance user satisfaction and ensure sustainable development of IoT-enhanced educational platforms. Future research could explore more advanced mathematical models and context-specific challenges to refine risk prioritization strategies.

5. A Fuzzy Logic Approach for Prioritizing Customer Retention Strategies in OTT Video Platforms

   Bincy Ann Babu, N. Elangovan, Kerena Anand, Jacob Joseph Kalapurackal

   Abstract

   The emergence of over-the-top (OTT) video platforms has significantly transformed the way people consume content and the entire media landscape. The significant growth of OTT video platforms in recent years, amidst fierce competition and changing consumer preferences, has posed a challenge for the platforms in retaining customers. Elevated churn rates in the OTT video platforms have prompted them to focus on customer retention. OTT platforms implemented different strategies to retain customers. The customer retention strategies are identified through a literature review and unstructured interviews with industry experts. This paper presents a novel approach to prioritize customer retention strategies using a fuzzy analytic hierarchy process (fuzzy AHP). The fuzzy AHP analysis results show that content strategy is the most significant, followed by pricing, customer experience, and platform extension. This paper provides actionable insights for OTT platform managers, helping them enhance user satisfaction and retain customers.

6. IoT Based Control Mechanism for Solar PV Water Pumps Employed for Water Distribution in Academic Campus

   Basanagouda Ronad, Sangamesh Goudappanavar, Basavaraj Hugar, Santosh Raikar, Santosh Kumbalavati, Kiran Bendigeri

   Abstract

   In this paper, IoT based control mechanism for solar water pumps employed for academic campus is presented. Basaveshwar Engineering College Bagalkote (BEC) (Latitude 16° 59′, Longitude 75° 59′) is using solar photovoltaic based water pumps, for water distribution in the campus. Three water pumps of 0.5, 1, and 2 HP are employed with 400, 900, and 1800 W panels respectively. The overhead tanks of the main building of the campus are filled by these pumps. The operators from the campus development department claimed the difficulty in operating the pumps for filling the overhead tanks. Due to variation in solar radiation it is difficult to monitor and manage physically, the different tanks with sufficient water availability. This issue not only results in water wastage by over spilling but also requires significant human effort and time for continuous observation for manual adjustments. In view of this, it is proposed to provide the IoT based solution for control mechanisms of SPV based water pumps. The proposed circuitry is successfully implemented and control of pumps is carried using mobile phones. Further, the system with 0.5 HP AC grid connected water pump is also additionally established to pump water during prolonged cloudy conditions. The economic analysis of the said water pumps for the last eight years was conducted. It is observed that the solar PV system is economical, very effective and also resulted in significant energy savings. Further, the implementation of the proposed IoT methodology resulted in the smooth operation of the pumps from the remote locations.

7. Enhanced Named Entity Recognition in Marathi News Articles Using Machine Learning Approach

   Nita V. Patil, Ajay S. Patil

   Abstract

   Named Entity Recognition (NER) is an essential component of natural language processing. Marathi is an under-resourced language used by significant worldwide users specially in Maharashtra. The primary objective of this study is to enhance NER development in Marathi news articles by applying the machine learning algorithm Hidden Markov Model (HMM) with a pattern matching approach. The NER system trains HMM on a manually annotated news corpus using the Viterbi decoding algorithm to assign the most probable named entity tags to words in the test dataset. The baseline NER system achieves a 63.6% named entity identification and a 72.95% classification accuracy. The NER system's identification and classification rate is improved to 82.07% and 86.84%, respectively, by integrating pattern matching techniques. This study sets a benchmark for Marathi NER development vitally used for natural language processing tasks such as machine translation, question answering, summarization, and information retrieval systems. The study presented in this paper is an important basic component that uses robust methodology, a detailed dataset preparation method, and comprehensive related research, ensuring the reliability and validity of the named entity recognition results.

8. Comparative Analysis and Hyperparameter Optimization of Machine Learning Models for Predicting Soil Fertility

   Bhaskar Vijayrao Patil, Deepali M. Gala, Bhakti Pawar, Kirti Khanna, Arnab Chakraborty, Gayathri Band

   Abstract

   This study presents a detailed comparative analysis of various ML models—Support Vector Machine (SVM), Random Forest (RF), Naive Bayes, KNN, Decision Tree (DT), and XGBoost—for predicting soil fertility based on critical soil characteristics. Using a comprehensive dataset that includes Nitrogen (N), Phosphorus (P), Potassium (K), temperature, pH, and humidity, the research applies standard preprocessing techniques and evaluates model performance through rigorous cross-validation and validation accuracy metrics. Among the models evaluated, Random Forest demonstrated the highest performance, which was further improved through hyperparameter tuning using GridSearchCV. The study also utilized SHAP values to gain insights into feature importance, revealing key factors such as Nitrogen content, pH level, and Potassium content as most influential in predicting soil fertility. The results highlight the effectiveness of model selection and optimization techniques in achieving high prediction accuracy. This research not only underscores the potential of ML to enhance precision farming practices but also provides valuable data-driven insights for improving agricultural sustainability. The findings contribute significantly to the advancement of precision agriculture by offering actionable strategies for soil management and fertility enhancement, thereby supporting the development of more resilient and sustainable agricultural systems. The study also leveraged SHAP values to gain insights into feature importance, identifying Nitrogen content, pH level, and Potassium content as the most influential factors in predicting soil fertility. These findings underscore the effectiveness of model selection and optimization techniques in achieving high prediction accuracy and highlight the potential of ML to improve precision farming practices. The research offers valuable data-driven insights for enhancing agricultural sustainability, contributing to the advancement of precision agriculture by providing actionable strategies for soil management and fertility enhancement. Future research should focus on expanding datasets to include diverse regions and additional features, exploring advanced ML techniques, and addressing real-world implementation challenges.

9. Unveiling Cybercrime Patterns in Kerala: A Machine Learning Approach

   M. P. Swapna, Devi Rajeev, J. Ramkumar, Saranya Chandran

   Abstract

   Security measures are vital in the present day and age, especially in regions such as Kerala, India, where the use of the internet and digital technology is rapidly growing hand in hand with new threats. This work therefore sought to establish the participants’ levels of awareness and experiences with cybercrime in Kerala through machine learning models. The research is based on the survey data of users demography (age, gender, education, urban environment/rural area) awareness scores, and personal experience of cybercrimes (phishing, malware attacks, identity theft, and online fraud). This entails creating and distributing a questionnaire to obtain the required data; the next steps include data cleaning aspects such as missing data imputation, creating dummy variables for the categorical variables, and scaling for numerical ones. To make predictions from the given data, the Decision Trees, Random Forests, Logistic Regression, K-means Clustering, and Support Vector Machines algorithms are used. These algorithms are chosen because they can effectively distinguish cybersecurity events and identify factors that might influence people to become cybercrime victims. Important factors to be noted are the correlation between the demographic data and the existing level of awareness regarding cybersecurity, the frequency rates for several types of cybercrimes, and the evaluation of the extant models of cybersecurity in the state of Kerala in India. As found, to overcome the issue, it is necessary to launch appropriate awareness programs and implement greatly needed cybersecurity initiatives, which should always be adapted to the demographics of the regions with high cyber risks and the ways people utilize digital technologies. Hence, the study provides government policymakers, cyber-security workers, and other interested researchers a better understanding about cybersecurity status and future prospects of Kerala, India, and other similar digital societies.

10. Cervical Spine Posture Monitoring using Flex and IMU Sensors with Long Short Term Memory Networks

    Arnav Gupta, Reetu Jain

    Abstract

    Neck pain or cervicalgia is the discomfort in or around the cervical spine which is the region of the spine located just below the head. One of the most common causes for neck pain is poor posture. Excessive use of electronic devices like computers, cell phones, etc aggravates the neck posture. This research focused on the development of a device that monitors cervical spine posture and provides feedback to the user to help them maintain a good neck posture. The device uses three flex sensors to measure neck bending and an MPU6050 IMU sensor for measuring acceleration and rotation in the neck movements. It employs an ERM motor to provide haptic feedback to the user about their neck posture. HC-05 Bluetooth modules have been used for wireless communication and a LiPo battery powers the device. A Bluetooth adapter, which contains a HC-05 and a USB-to-TTL converter, transmits sensor data to a computer running the Python script that feeds this data to a Long Short Term Memory (LSTM) Network to predict the user's neck posture type. The neck band containing the sensors has a diameter of 110 mm and height of 120 mm. It is made up of flexible Thermo Polyurethane and has a velcro for easy wear. An enclosure made up of Polylactic Acid with its dimensions as 56 mm × 56 mm × 58 mm houses all the components of the device including the vibration motor. The enclosure has a push button, ports for uploading code and charging the battery, and a hole for sensor wires coming from the neck band. The LSTM model was trained on a dataset containing 3400 data samples. The device uses Python to feed the sensor data to the trained LSTM model which classifies the neck postures into 2 categories, good and bad. The Python script also sends the posture type predictions back to the device via Bluetooth. Real-time sensor data for every user wearing the device is saved in CSV files along with the predicted posture labels. Based on the LSTM model predictions about the neck posture type received from the python script, the Arduino code running on the device controls the vibration motor and alerts the user with two vibrations for bad posture and no vibrations for good posture. The LSTM model's classification performance was evaluated using accuracy. Over 100 epochs, validation loss decreased from 70 to 28.51%, and validation accuracy increased from 40.59 to 95%. The accuracy obtained on the test set was 82%. User testing with 20 participants showed that the device accurately distinguished between good and bad neck postures with an overall accuracy of 79%, which varied by activity: 81% at a desk, 82% while walking, and 78% while sitting in different postures. User feedback indicated 75% found the device comfortable for extended wear, 85% found the neck band to be non-intrusive, 65% participants could easily set it up, and 40% reported minor discomfort from the vibration motor. Additionally, 80% felt the device feedback helped them correct their neck posture with 75% finding the vibration alert effective, and 70% agreeing that the usage of the device improved their neck posture awareness.

11. A Comprehensive Study on Enhancing Spatial Privacy: Adaptive Noise Integration in Point-Based and Set-Based Differential Privacy Approaches

    Mohammed Hasan Ahmed Abdullah, Nemi Chandra Rathore

    Abstract

    Although mobile apps and location-based services provide customers with until unheard-of convenience and personalizing, they seriously compromise their location privacy. In the framework of location-based services, this study tackles the crucial difficulty of harmonizing privacy protection with data utility. Including both point-based and set-based differential privacy strategies, we offer a thorough evaluation of several Location Privacy Preserving Approaches. We assess these LPPAs over a range of privacy budgets. We provide a new Polar mechanism with adaptive noise using location density to dynamically change noise levels, hence minimizing the trade-off between privacy and accuracy. We show that our proposed approach regularly beats baseline techniques in terms of spatial inaccuracy and root mean square error. Detailed empirical studies show that these cutting-edge methods preserve strong privacy protection and improve accuracy.

12. A Solution to Fuschian Differential Equation Having Five Singular Points

    Brajesh Shukla, Vinay Shukla, Sumit Malik, Sanjay Yadav

    Abstract

    Generalized Heun-type differential equation appears in the study of quantum systems, particularly in the calculation of energy levels and wave functions. A particular case of generalized Heun type differential equation is solved using a special approach called tridiagonal representation approach. The differential equation has five singularities: four regular and one irregular at infinity. The differential equation is solved in terms of Jacobi polynomials in a series. This series’ expansion coefficients fulfill a three-term recurrence relation, so forming a new class of orthogonal polynomials.

13. Development and Analysis of Wake Effect of Vertical-Axis Wind Turbine Generator on HAWTG Based on Experimental Method

    Sangamesh Y. Goudappanavar, Suresh H. Jangamshetti, Shreeshail Majjagi, Basanagouda F. Ronad

    Abstract

    In this paper, development of experimental setup to identify the wind velocity deficit due to wake effect of Vertical-Axis Wind Turbine (VAWT) on Horizontal-Axis Wind Turbine (HAWT) in Renewable Energy Research Laboratory Basaveshwar Engineering College, Bagalkote. Downstream turbines have reduced incident energy and momentum. A VAWT causes a wake behind it to extend linearly when a uniform wind strikes it. The free wind will experience a partial reduction in speed from V_up to V_down. The wake effect in which the incoming wind with speed V_up hits the blades of turbines and creates a wake cone. Downstream wind turbines inside the wake cone experience the velocity deficit in wind speed and extracts lesser power. Energy wake losses typically 5–20%. Results show that 42–23% on HAWT at a various distance. Performance results available from experimental setup of proposed work investigate as distance increases, the velocity deficit decreases, indicating that the wake effect diminishes with distance. The similarity between theoretical and experimental values suggests that the theoretical model is accurate in predicting the wake effects on a Horizontal-Axis Wind Turbine (HAWT) positioned downstream. This aims to estimate the turbulence that leads to extensive power output variations. Evaluation of the energy efficiency and power smoothing of wind turbines.

14. A Zero-Trust Security Paradigm for Banking Transactions: Strengthening Defences Against Financial Fraud and Cyberthreats

    Joy V Ramachandran, Rashmi Agarwal

    Abstract

    Banking transactions are increasingly problematic, with new vulnerabilities emerging daily. Data shows nearly 47% of reported fraud cases were card related, 18% involved mobile fraud, 11% were attributed to merchant fraud, 13% were digital fraud, 6% concerned virtual currency fraud, 4% were related to cheque fraud, and 25% of countries reported issues with lost or stolen cards. Additionally, 20% of countries reported account takeover fraud, and 15% experienced identity spoofing. To combat these challenges, individuals, industries, organizations, and financial institutions worldwide need a banking transaction system that minimizes fraud, is user-friendly, and robust. This paper proposes integrating existing technologies with a zero-trust security model to address contemporary banking fraud issues. Specifically, it introduces a device-specific trust mechanism that generates a dynamic One-Time Password (OTP) only if a secure handshake occurs between the banking system and the device each time a customer engages in banking activities, whether at an Automated Teller Machine (ATM), or during online purchases, through net banking transfers, or at a Point-of-Sale (POS) system. This paper explores a novel banking transaction security mechanism involving dynamic Quick Response (QR) codes for transaction authentication. Upon card insertion, the ATM generates a QR code based on a hashed master password, Integrated Circuit Card Identification Number (ICCID), and Mobile Station Equipment Identity (IMEI) from the device registration database. To ensure the QR code is unique for each transaction, a seed value representing the elapsed time of the day is incorporated. The customer scans the QR code with their mobile banking application, generating a matching hash value. Upon a successful match, the mobile banking app displays a four-digit OTP derived from the hash value, which the customer uses to complete the ATM transaction. The paper details the algorithm’s functionality and unique OTP generation approach and addresses potential challenges related to timing, synchronization, and implementation complexity.

15. A Fuzzy Analytic Hierarchy Process Approach to Prioritize Governance Factors Within Environmental Social Governance Factors in the BFSI Sector

    Geeta Maladkar, B. Suresha, Aparna Hawaldar, R. Anuradha

    Abstract

    The banking, financial services, and insurance industry is progressively incorporating environmental, social, and governance considerations to promote sustainable and responsible practices. This research concentrates on prioritizing the governance components within the ESG framework, highlighting their crucial influence on the BFSI sector. Utilizing an Analytic Hierarchy Process (AHP) combined with fuzzy logic, we systematically evaluate and rank governance factors for a robust decision-making framework based on expert insights from the BFSI sector. The AHP methodology structured and weighted governance criteria, while fuzzy logic addressed uncertainty and subjectivity in expert judgments. The findings identify the most crucial governance factors, offering valuable insights for stakeholders to strengthen governance practices and align them with ESG goals. This study contributes to sustainable finance literature by providing a comprehensive methodology for prioritizing governance aspects, promoting resilience and ethical standards in the BFSI sector.

16. AI-Based Risk Analysis and Preemptive Prediction of Rheumatoid Arthritis

    Y. K. Anupama, Ishika Jain, Namrata B Hakari, Shreya Gunti, Sidhanti Patil

    Abstract

    Rheumatoid Arthritis (RHA) presents a significant challenge in modern medicine. Despite advancements in modern medicine and preventive healthcare, Rheumatoid Arthritis (RHA) remains an unpredictable and largely undetectable disease in its early stages. This paper proposes a RHA prediction model framework to address this challenge. Our goal is twofold: to enable preemptive prediction of RHA and to empower patients diagnosed with RHA. The proposed solution is a comprehensive RHA management platform with various features, including preliminary risk assessment, risk analysis module, visual DAS 28 calculator, community support, and awareness. The system employs a comprehensive pipeline that includes image preprocessing, feature extraction, and an RHA prediction model, achieving improvisation in terms of accuracy.

17. An Evidence-Based Blockchain Digital Literacy Model for Sustainable Smart Village: A Step Toward Minimizing Rural–Urban Digital Divide

    Syed Imtiyaz Hassan

    Abstract

    Sustainable development is one of the major concerns of the world these days. Various efforts in different dimensions by different governments and organizations have already been initiated to address the concerns of sustainability. Modern emerging disruptive technologies are also employed for digital services of smart cities and smart villages that reach the masses. However, statistics suggest that there is still a gap between rural and urban communities due to various factors. The lack of digital literacy among rural communities is one of the contributing factors leading to rural–urban digital divide. The present research is, therefore, an attempt to examine the world perspective of the rural–urban digital divide, to identify different levels of digital divide, to discuss various digital literacy steps that may be taken to reduce such digital divide, and to explore the role of blockchain in that context. Further, a blockchain-based digital literacy model for sustainable smart villages is proposed for the purpose of minimizing rural–urban digital divide. The potential benefits of the proposed evidence-based digital literacy model are real-time monitoring and control, smart contracts enforcement, literacy level-based personalization, and motivation through incentives, leading to transparency, accountability, and improved service delivery.

18. Prepartum Prediction of Cephalopelvic Disproportion Based on Maternal Anthropometry, Classification of Shape of Pelvic Bone and Head Circumference of the Foetus

    P. Sandhya, Anik Bhaumik, R. Srivats, V. Kalyanasundaram, Amogh Singh

    Abstract

    Cephalopelvic Disproportion (CPD) is a condition where the fetal head or body is too large to pass through the mother’s pelvis, affecting about 1 in 250 pregnancies. This complication often leads to emergency Cesarean sections, especially in rural areas with limited emergency care facilities. Traditional prediction methods, based primarily on maternal anthropometric measurements, achieve a limited accuracy of about 24%, prompting a need for more reliable approaches. This study proposes an innovative method that integrates maternal anthropometry, pelvic bone shape classification via MRI, and fetal head circumference to enhance CPD prediction accuracy. Using a dataset of 500 DICOM images, pelvic shapes were categorized into Gynecoid (lower CPD risk) and non-Gynecoid (higher CPD risk) types. Non-Gynecoid cases underwent further analysis of fetal head circumference using ultrasound. Advanced machine learning and deep learning techniques, including Support Vector Machine (SVM), K-Nearest Neighbors (KNN), and Convolutional Neural Networks (CNN), were employed to classify the images, achieving accuracy rates of up to 97.62%. These findings suggest that combining multiple diagnostic parameters provides a more robust pre-labor CPD prediction tool, potentially reducing emergency interventions and improving maternal and fetal outcomes, particularly in settings with limited access to advanced medical care. Future work will focus on expanding the dataset and refining the algorithms to further enhance prediction accuracy. This approach paves the way for early, non-invasive intervention strategies, offering significant benefits for both mothers and their newborns, particularly in settings with limited access to advanced medical care.