Nana Yaw Asabere
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Nana Yaw Asabere has a total of 80 co-authors. Here are the top co-authors by number of shared publications:
- Amevi Acakpovi
Accra Technical University
12 shared publications - Gare Lawson
Accra Technical University
4 shared publications - Joseph Agyiri
Accra Technical University, Ghana
4 shared publications - Wisdom Kwawu Torgby
Accra Technical University, Ghana
4 shared publications Eric Amoako
Accra Technical University, Ghana
3 shared publicationsGare Lawson
3 shared publications
Latest Publications
- Open Access
- 20 May 2026
- Published by Springer Nature in Journal of Electrical Systems and Information Technology
- Vol. 13 (1)
Monitoring and analysing student attentiveness in traditional classrooms through computational intelligence
In traditional face-to-face mode of education, keeping an eye on students’ concentration is still quite difficult, especially in large or packed lecture halls/classrooms where teachers are unable to consistently monitor minor behavioural indications of disengagement. In this paper, we propose a framework—Student Attentive Analysis using Computational Intelligence ( SAA-CI ). SAA-CI is a real-time, privacy-conscious framework that uses non-intrusive still-image sequences to evaluate students’ attentiveness using computer vision and deep learning. SAA-CI takes periodic snapshots and uses OpenCV and Dlib to extract important behavioural indications such as gaze direction, head attitude, eyelid movement, yawning, and facial expressions, in contrast to current systems that rely on continuous video or invasive sensors. These traits are recorded as temporal vectors and classified using a Long Short-Term Memory (LSTM) network that has been improved with an attention mechanism. Through an instructor dashboard and adaptive intervention engine, SAA-CI provides actionable pedagogical insights and binary attentiveness forecasts (Attentive/Not Attentive). Scientific benchmarking experiments on the DAiSEE dataset and a special SAA-CI -Local classroom dataset show that SAA-CI performs better than current baseline techniques, with 92.6% accuracy, a 9.4% gain due to the attention mechanism, and additional gains from data augmentation. The findings illustrate that SAA-CI maintains high dependability across a range of classroom behaviours, functions effectively in real time, and protects student privacy with optional anonymization and low data retention. By providing a scalable, morally sound, and pedagogically significant attentiveness monitoring solution for contemporary educational settings, SAA-CI improves intelligent classroom analytics.
- 20 May 2026
- Published by Springer Nature in Journal of Electrical Systems and Information Technology
- Vol. 13 (1)
Open AccessMonitoring and analysing student attentiveness in traditional classrooms through computational intelligence
In traditional face-to-face mode of education, keeping an eye on students’ concentration is still quite difficult, especially in large or packed lecture halls/classrooms where teachers are unable to consistently monitor minor behavioural indications of disengagement. In this paper, we propose a framework—Student Attentive Analysis using Computational Intelligence ( SAA-CI ). SAA-CI is a real-time, privacy-conscious framework that uses non-intrusive still-image sequences to evaluate students’ attentiveness using computer vision and deep learning. SAA-CI takes periodic snapshots and uses OpenCV and Dlib to extract important behavioural indications such as gaze direction, head attitude, eyelid movement, yawning, and facial expressions, in contrast to current systems that rely on continuous video or invasive sensors. These traits are recorded as temporal vectors and classified using a Long Short-Term Memory (LSTM) network that has been improved with an attention mechanism. Through an instructor dashboard and adaptive intervention engine, SAA-CI provides actionable pedagogical insights and binary attentiveness forecasts (Attentive/Not Attentive). Scientific benchmarking experiments on the DAiSEE dataset and a special SAA-CI -Local classroom dataset show that SAA-CI performs better than current baseline techniques, with 92.6% accuracy, a 9.4% gain due to the attention mechanism, and additional gains from data augmentation. The findings illustrate that SAA-CI maintains high dependability across a range of classroom behaviours, functions effectively in real time, and protects student privacy with optional anonymization and low data retention. By providing a scalable, morally sound, and pedagogically significant attentiveness monitoring solution for contemporary educational settings, SAA-CI improves intelligent classroom analytics.
- Open Access
- 13 April 2026
- Published by Springer Nature in Discover Data
- Vol. 4 (1)
Geospatial-based location-aware recommendation of banking services in the world of big data
- 13 April 2026
- Published by Springer Nature in Discover Data
- Vol. 4 (1)
Open AccessGeospatial-based location-aware recommendation of banking services in the world of big data
- 24 February 2026
- Published by SAGE Publications in Journal of Ambient Intelligence and Smart Environments
- Vol. 18 (2) , 249-276
Geo-Lecture : Exploiting location-based recommendation in the context of lecture attendance tracking using a geographical information system (GIS)
To ensure productivity in the higher education sector, tracking lecture attendance of instructors and students is vital. Low attendance often leads to negative consequences in terms of optimal output for both individuals and the institution. Traditional methods of locating lecture rooms and tracking attendance are inefficient and error-prone. In this article, we propose Geo-Lecture , a location-based recommender algorithm which leverages a geographical information system (GIS). Using student and location data from a Technical University in Ghana, Geo-Lecture provides accurate, and real-time lecture attendance tracking. Evaluation results demonstrate that Geo-Lecture significantly outperforms contemporary methods with the highest precision of 0.31, recall of 0.34, and an F1-score of 0.32. Additionally, the mean absolute error (MAE) and Normalized MAE both illustrate results of 0.69 and 0.17, respectively, which are the lowest comparatively. These results validate Geo-Lecture 's superiority in terms of accuracy, reliability, and usability in comparison to existing approaches.
- 24 February 2026
- Published by SAGE Publications in Journal of Ambient Intelligence and Smart Environments
- Vol. 18 (2) , 249-276
Geo-Lecture : Exploiting location-based recommendation in the context of lecture attendance tracking using a geographical information system (GIS)
To ensure productivity in the higher education sector, tracking lecture attendance of instructors and students is vital. Low attendance often leads to negative consequences in terms of optimal output for both individuals and the institution. Traditional methods of locating lecture rooms and tracking attendance are inefficient and error-prone. In this article, we propose Geo-Lecture , a location-based recommender algorithm which leverages a geographical information system (GIS). Using student and location data from a Technical University in Ghana, Geo-Lecture provides accurate, and real-time lecture attendance tracking. Evaluation results demonstrate that Geo-Lecture significantly outperforms contemporary methods with the highest precision of 0.31, recall of 0.34, and an F1-score of 0.32. Additionally, the mean absolute error (MAE) and Normalized MAE both illustrate results of 0.69 and 0.17, respectively, which are the lowest comparatively. These results validate Geo-Lecture 's superiority in terms of accuracy, reliability, and usability in comparison to existing approaches.
- Open Access
- 2 February 2026
- Published by IntechOpen
Perspective Chapter: P-Tracer – Improving Disease Surveillance and Monitoring through Geospatial Technology
Although efficient surveillance and contact tracing are essential for being prepared for epidemics, the scalability, interoperability, and dependability of present systems face challenges. Consequently, in this chapter, we proposed Patient Tracer (P-Tracer), a digital health platform that runs on Field Data Collection Interfaces and includes modules for analytics, synchronization, data collection, contact tracing, and notifications. The system uses a dashboard for timely monitoring and an offline-first strategy to facilitate data collecting in places with poor connectivity. It also synchronizes in the background to a central server. Benchmarking scientific experimental analysis compared with other contemporary methods showed that demographics, symptoms, vital signs, and geolocation data could be reliably recorded, and that data synchronization between devices was more proficient and reliable in the case of P-Tracer.
- 2 February 2026
- Published by IntechOpen
Open AccessPerspective Chapter: P-Tracer – Improving Disease Surveillance and Monitoring through Geospatial Technology
Although efficient surveillance and contact tracing are essential for being prepared for epidemics, the scalability, interoperability, and dependability of present systems face challenges. Consequently, in this chapter, we proposed Patient Tracer (P-Tracer), a digital health platform that runs on Field Data Collection Interfaces and includes modules for analytics, synchronization, data collection, contact tracing, and notifications. The system uses a dashboard for timely monitoring and an offline-first strategy to facilitate data collecting in places with poor connectivity. It also synchronizes in the background to a central server. Benchmarking scientific experimental analysis compared with other contemporary methods showed that demographics, symptoms, vital signs, and geolocation data could be reliably recorded, and that data synchronization between devices was more proficient and reliable in the case of P-Tracer.
- Open Access
- 1 January 2026
- Published by African Journals Online (AJOL) in African Journal of Social Issues
- Vol. 9 (1) , 1322-1351
Comparative machine-learning prediction of concrete compressive strength: evaluating catboost and ensemble models
This study develops and evaluates machine-learning (ML) models for predicting concrete compressive strength and deploys the best-performing model as a prototype web-based estimation tool. Using 1,030 concrete-mixture records and eight mix-design variables, Multiple Linear Regression (MLR) was benchmarked against Decision Tree, Random Forest, Gradient Boosting, SVR, ANN, CatBoost, and a Stacked Ensemble. CatBoost, Random Forest, Gradient Boosting, SVR, and ANN were tuned via grid search with 10-fold cross-validation on an 80% training partition, with the remaining 20% held out as a single test split for final evaluation; the Decision Tree and Stacked Ensemble were not separately tuned. On this held-out test partition, all ML models produced higher R² than MLR (R² = 0.628). CatBoost achieved the strongest test performance (R² = 0.934; RMSE = 4.113 MPa), followed by the Stacked Ensemble (R² = 0.894) and Random Forest (R² = 0.882); these single-split comparisons have not been confirmed with repeated or nested cross-validation or formal statistical testing. CatBoost was retrained on the full 1,030-observation dataset and deployed as a prototype web-based calculator for rapid preliminary strength estimation. The tool supports, rather than replaces, laboratory testing and engineering judgement. With further validation, prediction-assisted mix design could facilitate exploration of lower-cement and higher supplementary cementitious material mixes, potentially improving material efficiency and reducing embodied carbon, although environmental and cost outcomes were not directly assessed.
- 1 January 2026
- Published by African Journals Online (AJOL) in African Journal of Social Issues
- Vol. 9 (1) , 1322-1351
Open AccessComparative machine-learning prediction of concrete compressive strength: evaluating catboost and ensemble models
This study develops and evaluates machine-learning (ML) models for predicting concrete compressive strength and deploys the best-performing model as a prototype web-based estimation tool. Using 1,030 concrete-mixture records and eight mix-design variables, Multiple Linear Regression (MLR) was benchmarked against Decision Tree, Random Forest, Gradient Boosting, SVR, ANN, CatBoost, and a Stacked Ensemble. CatBoost, Random Forest, Gradient Boosting, SVR, and ANN were tuned via grid search with 10-fold cross-validation on an 80% training partition, with the remaining 20% held out as a single test split for final evaluation; the Decision Tree and Stacked Ensemble were not separately tuned. On this held-out test partition, all ML models produced higher R² than MLR (R² = 0.628). CatBoost achieved the strongest test performance (R² = 0.934; RMSE = 4.113 MPa), followed by the Stacked Ensemble (R² = 0.894) and Random Forest (R² = 0.882); these single-split comparisons have not been confirmed with repeated or nested cross-validation or formal statistical testing. CatBoost was retrained on the full 1,030-observation dataset and deployed as a prototype web-based calculator for rapid preliminary strength estimation. The tool supports, rather than replaces, laboratory testing and engineering judgement. With further validation, prediction-assisted mix design could facilitate exploration of lower-cement and higher supplementary cementitious material mixes, potentially improving material efficiency and reducing embodied carbon, although environmental and cost outcomes were not directly assessed.
- 10 November 2025
- Published by Taylor & Francis in African Journal of Science, Technology, Innovation and Development
- Vol. 17 (7) , 1064-1079
GeoHostel : Location-based recommendation of hostels in the context of higher education through geospatial technology
- 10 November 2025
- Published by Taylor & Francis in African Journal of Science, Technology, Innovation and Development
- Vol. 17 (7) , 1064-1079
GeoHostel : Location-based recommendation of hostels in the context of higher education through geospatial technology
- 6 April 2025
- Published by Elsevier in Engineering Applications of Artificial Intelligence
- Vol. 151
An integrated multi-scale context-aware network for efficient desnowing
- 6 April 2025
- Published by Elsevier in Engineering Applications of Artificial Intelligence
- Vol. 151
An integrated multi-scale context-aware network for efficient desnowing
- Open Access
- 1 January 2025
- Published by Wiley in Advances in Public Health
- Vol. 2025 (1)
Improving Counseling Sessions Through an Interactive Web‐Based Application in the Context of Higher Education
The existing face‐to‐face (F2F) method of counseling in most of Ghana’s public tertiary institutions is accompanied by difficulties and challenges, such as stigmatization, depression among students, and limited counseling access. Our research methodology utilized a questionnaire research instrument as a pathway for substantiating the development of our proposed computerized software solution. A total of 434 students and 10 counselors from public universities in Accra responded to the questionnaires. To corroborate the responses received and provide an appropriate solution, we proposed, developed, and implemented an online (web‐based) interactive counseling application for relevant stakeholders in the case study universities, through a Design Science Research (DSR) approach. Implementation results through computerized evaluation showed that our proposed online counseling system is highly beneficial, and it provides optimum solutions to the challenges.
- 1 January 2025
- Published by Wiley in Advances in Public Health
- Vol. 2025 (1)
Open AccessImproving Counseling Sessions Through an Interactive Web‐Based Application in the Context of Higher Education
The existing face‐to‐face (F2F) method of counseling in most of Ghana’s public tertiary institutions is accompanied by difficulties and challenges, such as stigmatization, depression among students, and limited counseling access. Our research methodology utilized a questionnaire research instrument as a pathway for substantiating the development of our proposed computerized software solution. A total of 434 students and 10 counselors from public universities in Accra responded to the questionnaires. To corroborate the responses received and provide an appropriate solution, we proposed, developed, and implemented an online (web‐based) interactive counseling application for relevant stakeholders in the case study universities, through a Design Science Research (DSR) approach. Implementation results through computerized evaluation showed that our proposed online counseling system is highly beneficial, and it provides optimum solutions to the challenges.
- Open Access
- 12 July 2024
- Published by Wiley
REAL-TIME HUMAN DETECTION AND COUNTING OF CUSTOMERS IN THE BANKING INDUSTRY IN GHANA: A COMPUTER VISION AND DEEP LEARNING APPROACH TO SAFETY AND SECURITY
The Guarantee Trusted Bank (GT Bank) at Opera Square, Accra, currently faces challenges in accurately counting and monitoring customers entering and exiting the bank premises, leading to operational inefficiencies and safety concerns. Additionally, the current system for counting and monitoring customers through both a CCTV and physical approach in the bank is prone to errors and cannot collect and store data for future planning. Security personnel also lack knowledge of the exact number of daily visitors and struggle to determine the exact timing of incidents during working and closing hours. This study aims to explore the weaknesses of the current detection and counting system at the GT bank and proposes solutions to address these issues. The primary solution involves the implementation of a computerized human detection and counting system, designed to accurately detect and count individuals who enter and exit the GT bank’s branch premises. To achieve this goal, a qualitative methodology is employed. Our qualitative method involved interviewing participants using relevant open-ended questions, which motivated them to express their thoughts and views openly with no limitations. The analysis of quantitative data illustrates the difficulties faced by the current system in detecting and counting customers in the bank. Furthermore, the research participants are willing to accept our proposed system, which would improve the current counting and monitoring system used by the bank. Our proposed system utilizes computer vision and deep learning algorithms through the use of a single-shot multi-box detector (SSD) to identify human figures in video feeds. By introducing the proposed system within the GT bank, it can improve the overall customer experience. One implication for decision-makers involves the designation of specialized personnel responsible for system maintenance and troubleshooting, thereby assisting in routine checks and upkeep to prevent system downtime and maintain continuous operation.
- 12 July 2024
- Published by Wiley
Open AccessREAL-TIME HUMAN DETECTION AND COUNTING OF CUSTOMERS IN THE BANKING INDUSTRY IN GHANA: A COMPUTER VISION AND DEEP LEARNING APPROACH TO SAFETY AND SECURITY
The Guarantee Trusted Bank (GT Bank) at Opera Square, Accra, currently faces challenges in accurately counting and monitoring customers entering and exiting the bank premises, leading to operational inefficiencies and safety concerns. Additionally, the current system for counting and monitoring customers through both a CCTV and physical approach in the bank is prone to errors and cannot collect and store data for future planning. Security personnel also lack knowledge of the exact number of daily visitors and struggle to determine the exact timing of incidents during working and closing hours. This study aims to explore the weaknesses of the current detection and counting system at the GT bank and proposes solutions to address these issues. The primary solution involves the implementation of a computerized human detection and counting system, designed to accurately detect and count individuals who enter and exit the GT bank’s branch premises. To achieve this goal, a qualitative methodology is employed. Our qualitative method involved interviewing participants using relevant open-ended questions, which motivated them to express their thoughts and views openly with no limitations. The analysis of quantitative data illustrates the difficulties faced by the current system in detecting and counting customers in the bank. Furthermore, the research participants are willing to accept our proposed system, which would improve the current counting and monitoring system used by the bank. Our proposed system utilizes computer vision and deep learning algorithms through the use of a single-shot multi-box detector (SSD) to identify human figures in video feeds. By introducing the proposed system within the GT bank, it can improve the overall customer experience. One implication for decision-makers involves the designation of specialized personnel responsible for system maintenance and troubleshooting, thereby assisting in routine checks and upkeep to prevent system downtime and maintain continuous operation.
- Open Access
- 1 January 2024
- Published by Wiley in Human Behavior and Emerging Technologies
- Vol. 2024 (1)
Geo-Insurance: Improving Big Data Challenges in the Context of Insurance Services Using a Geographical Information System (GIS)
Both large and small information flows can have a significant impact on how consumers obtain trustworthy financial information, ultimately leading to an improvement in their daily lives when they interact dynamically with local geographic conditions. In economies that face both geographical and socioeconomic challenges, such as those in Africa, this kind of context is crucial. Large information flows provide significant issues such as big data challenges in the insurance sector, which calls for robust, demand‐driven, and adaptive innovation solutions. In this paper, we present a geographic information system (GIS)–based location‐aware recommender algorithm, called Geo-Insurance. Using some selected insurance companies in Accra, Ghana, as a point of view for location and customer data, our proposed Geo-Insurance solution addresses the big data challenges of customers finding the closest insurance companies with specific services through a web‐based map created using a geodatabase file, ArcCatalog, and ArcGIS (among others). We conducted a series of benchmarking experiments. Our evaluation results show that Geo-Insurance performs better than other contemporary methods in terms of F‐measure (F1), recall (R), precision (P), mean absolute error (MAE), and normalized MAE (NMAE).
- 1 January 2024
- Published by Wiley in Human Behavior and Emerging Technologies
- Vol. 2024 (1)
Open AccessGeo-Insurance: Improving Big Data Challenges in the Context of Insurance Services Using a Geographical Information System (GIS)
Both large and small information flows can have a significant impact on how consumers obtain trustworthy financial information, ultimately leading to an improvement in their daily lives when they interact dynamically with local geographic conditions. In economies that face both geographical and socioeconomic challenges, such as those in Africa, this kind of context is crucial. Large information flows provide significant issues such as big data challenges in the insurance sector, which calls for robust, demand‐driven, and adaptive innovation solutions. In this paper, we present a geographic information system (GIS)–based location‐aware recommender algorithm, called Geo-Insurance. Using some selected insurance companies in Accra, Ghana, as a point of view for location and customer data, our proposed Geo-Insurance solution addresses the big data challenges of customers finding the closest insurance companies with specific services through a web‐based map created using a geodatabase file, ArcCatalog, and ArcGIS (among others). We conducted a series of benchmarking experiments. Our evaluation results show that Geo-Insurance performs better than other contemporary methods in terms of F‐measure (F1), recall (R), precision (P), mean absolute error (MAE), and normalized MAE (NMAE).