Ashesi Institutional Repository
Welcome to the Ashesi Institutional Repository (AIR), an archive for preserving and sharing scholarly work at Ashesi University.
Contributors to the repository ensure that their scholarly and creative work is preserved, indexed and showcased for a global audience. Students who produce strong research work have the privilege of getting their work published on AIR. Ashesi Community members can make submissions through warrenlibrary@ashesi.edu.gh.
The repository is organized in "Communities" that group publications by department and/or subject. You can browse the "Community" by Collection, Author, Title, or Issue Date. Alternatively, a search function enables easy access to information.
Communities
Select a community to browse its collections.
- Faculty and student publications on arts, social sciences and humanities subjects
- Faculty and student publications on STEM subjects
- An archive of all university publications
Recent Submissions
Use of Machine Learning for the Optimization of Genetic Circuits in Synthetic Biology: Focusing On Promoter Prediction for Gene Expression in Escherichia Coli
(2024-08) Ineza, Nice Cailie
This study explores the application of Machine Learning (ML) in optimizing genetic circuit design in Synthetic Biology, particularly focusing on predicting specific promoter for gene expression in Escherichia Coli. Despite the potential of ML, current methods rely heavily on trial-and-error, which is inefficient and costly. The research employs various ML models, including Genetic Algorithms, Support Vector Machines, and Neural Networks, alongside computational algorithms like Boyer-Moore-Horspool’s algorithm, to predict promoter efficacy and identify optimal promoter-gene configurations. Utilizing datasets from databases such as RegulonDB, EcoCyc, and PRODORIC, the study validates its findings through a combination of literature cross-checks and model performance metrics. The resulting model achieved a high accuracy in predicting promoter efficacy, with a 92% success rate in identifying optimal configurations. The findings suggest that incorporating transcription unit data significantly improves prediction accuracy, demonstrating the potential of ML in advancing synthetic biology towards more precise and efficient genetic circuit design.
Enhancing Retail Decision-Making Through Data Mining Techniques
(Ashesi University, 2024-08) Lartey, Ann-Vanessa
In an era where a deep understanding of consumer behaviour increasingly drives retail decision-making, this research explores how data mining techniques can transform the retail sector. According to [3], millions of transactions are made, and data is generated and stored in the software of those retailers, waiting for further analysis to become useful for decision-making. This thesis ventures beyond traditional analysis, employing advanced algorithms to examine transactional data from a selected supermarket. The research aims to reveal hidden patterns in supermarket transactional data. The main objective is to collect, pre-process, and analyse supermarket transactional data. This thesis employs algorithms such as Apriori, FP-growth, and NOVA classification. The Apriori algorithm revealed frequent co-purchases and significant associations between products. NOVA classification provided insights into the processing levels of purchased foods, revealing a predominance of ultra-processed foods, which has implications for consumer health and marketing strategies. The study concludes that data mining techniques can significantly enhance retail decision-making by providing actionable insights into consumer behaviour in Africa.
Intelligent Energy Management in Smart Buildings Using AIoT
(Ashesi University, 2024-08) Ampomah-Asiedu, Amma Oforiwaa; Buntugu, Wepea Adamwaba
The escalating national demand for energy and the need for the proper implementation of sustainable consumption practices have prompted the investigation and development of an intelligent energy management system using AIoT. Rapid advancements have provided a wide variety of technological solutions to prominent issues facing society. However, sustainable energy distribution remains a significant concern; one made more pressing by the climate change crisis. This paper explores the implementation of a Soft Actor Critic model for energy cost reduction while maintaining occupant comfort to optimize energy consumption in a smart building. This will form part of a larger AIoT system integrating renewable energy and data collection via MQTT. The system seeks to evaluate the effectiveness of intelligent energy management in providing sustainable financial benefits to the individual customer and a sustainable power economy to the Ghanaian market at large. Additionally, the system integrates a Time of Use pricing model to suggest prime energy usage periods to the user for the encouragement of off-peak hour consumption, suggesting TOU implementation in Ghana. Experimentation results suggest a SAC agent-controlled single family house would benefit from cost savings while reducing the emissions caused as a result of the pattern of electricity demand from the house. This integration not only improves convenience and energy optimization but also provides a financial incentive. Incorporating AIoT with reinforcement learning improves system adaptability and offers a scalable solution for applications in future smart grids.
Medlink: Revolutionizing Emergency Healthcare Drug Accessibility in Ghana Using a Novel Dynamic Geofencing Algorithm
(Ashesi University, 2024-08) Quarshie, Thomas Kojo Yesu
The inability to quickly access essential medications during emergencies can lead to tragic outcomes. In Ghana, this issue is particularly acute due to delayed arrival at hospitals and set back in locating pharmacies stocked with needed drugs. To address this critical problem lingering in our health sector, this study introduces a groundbreaking dynamic geofencing algorithm. While traditional static geofencing has been used extensively to support geriatric patients with Alzheimer and Dementia by sending notifications based on predefined zones to caregivers, little research has been done on dynamically defining geofences to locate pharmacies with needed medications. This research focuses on developing a dynamic geofencing algorithm that converts geographic coordinates into a cartesian system for improved efficiency in locating closest pharmacies stocked with the needed medications. Our approach leverages HCI principles, such as context-aware computing which enables the system to intelligently adapt to users’ real-time locations and needs, significantly reducing search time and effort. Through extensive experimentation and analysis, the proposed algorithm has demonstrated a substantial reduction in time required to locate pharmacies stocked with needed medication as compared to the traditional approach of moving from one pharmacy to the next. This study underscores the transformative potential of dynamic geofencing in cutting down delays in accessing emergency healthcare in Ghana. By providing reliable, real-time information on the availability of medications in nearby pharmacies, this research paves the way for timely access to emergency healthcare, ultimately improving health outcomes in resource constrained settings.
Wanderer:an Ai-Powered Career Readiness and Evaluation Platform
(Ashesi University, 2024-08) Boateng, Simon Junior; Wahab, Abdul-Abass
The job market post-pandemic has been very competitive, especially in the STEM field. This demands that students equip themselves with the necessary knowledge and skills to perform well during interviews to guarantee employment. Unfortunately, most students lack the proper guidance and resources to prepare for these interviews effectively. Many struggles with the practical application of theoretical concepts, problem-solving techniques, and the ability to demonstrate their knowledge and articulate their thoughts during technical interviews. Several companies have created mock interview platforms to assist students and professionals during their interviews. Platforms like Pramp and Leetcode allow students to practice coding problems, engage in peer-to-peer mock interviews but they fall short on providing personalized feedback based on the candidate’s performance. The proposed solution aims to provide students with personalized mock interview experiences. Wanderer streamlines the interview process by utilizing an AI (Artificial Intelligence) agent that carefully guides the user throughout the session by giving hints, suggestions, and customized feedback on their performance during and after the interview session. The successful implementation of the proposed solution would enable students to practice and be more conversant with the technical interview environment and reduce their chances of failure. Ultimately, the proposed solution would be an asset to students and professionals alike who want to practice and improve their interviewing knowledge and skills.