Delali Kwasi Dake

Mr. Delali Kwasi Dake

orcid

Research Focus & Metrics

Academic Career Length

6Years (2021-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
20 October 2022 in Education and Information Technologies
Verified
Journal Article
Closed access
PDF file
DDDelali Kwasi Dake
Delali Kwasi Dake
EGEsther Gyimah
Esther Gyimah

29 June 2021 in Technologies
Verified
Journal Article
Open access
PDF file
DDDelali Kwasi Dake
Delali Kwasi Dake
JGJames Gadze
James Gadze
GKGriffith Klogo
Griffith Klogo
Henry Nunoo-MensahHenry Nunoo-Mensah
Henry Nunoo-Mensah
The fast emergence of IoT devices and its accompanying big and complex data has necessitated a shift from the traditional networking architecture to software-defined networks (SDNs) in recent times. Routing optimization and DDoS protection in the network has become a necessity for mobile network operators in maintaining a good QoS and QoE for customers. Inspired by the recent advancement in Machine Learning and Deep Reinforcement Learning (DRL), we propose a novel MADDPG integrated Multiagent framework in SDN for efficient multipath routing optimization and malicious DDoS traffic detection and prevention in the network. The two MARL agents cooperate within the same environment to accomplish network optimization task within a shorter time. The state, action, and reward of the proposed framework were further modelled mathematically using the Markov Decision Process (MDP) and later integrated into the MADDPG algorithm. We compared the proposed MADDPG-based framework to DDPG for network metrics: delay, jitter, packet loss rate, bandwidth usage, and intrusion detection. The results show a significant improvement in network metrics with the two agents.

2 November 2022 in Mobile Information Systems
Verified
Journal Article
Open access
PDF file
DDDelali Kwasi Dake
Delali Kwasi Dake
CBCharles Buabeng-Andoh
Charles Buabeng-Andoh
Recently, students dropping out of school at the tertiary level without prior notice or permission has intrigued deep concern among academic authorities, instructors, and counsellors. It has therefore become necessary to understand factors that lead to high attrition rates among learners and identify at-risk students for urgent academic counselling. In providing a proactive response to learner attrition, the study deployed a machine learning algorithm with high model accuracy to predict students’ drop-out rates and identify dominant attributes that affect learner attrition and retention. An attrition model was built and validated among support vector machine, decision tree, multilayer perceptron, and random forest algorithms. The machine learning algorithms were tested for accuracy, precision, recall, F-measure, and ROC using the 10-fold and the 5-fold comparative cross-validation techniques. In addition to the cross-validation technique, the chi-square feature selection mechanism was implemented to understand the algorithms’ training time and accuracy. The random forest emerged as the best-performing algorithm, with an accuracy of 70.98% and 69.74% for the 10-fold and the 5-fold cross-validation implementations, respectively.

19 December 2025 in Discover Computing
Verified
Journal Article
Open access
PDF file
EOElijah Ofori
Elijah Ofori
DDDelali Kwasi Dake
Delali Kwasi Dake
Accurate prediction of student performance is essential for facilitating early interventions and enhancing academic results. However, interpretability is lacking in traditional machine learning approaches, which frequently struggle to capture implicit behavioural and learning dependencies within educational data. The study utilizes a publicly available dataset from the Kaggle repository, containing student records withfeatures related to academic, demographic, and behavioural attributes. The dataset was standardized using z-score normalization. Four deep learning models were developed based on their effectiveness in processing sequential data: LSTM, CNN, RNN, and a hybrid model combining LSTM and Transformer architectures (LSTM-Transformer). To ensure model transparency, we integrated Explainable Artificial Intelligence (XAI) techniques such as SHapley Additive exPlanations (SHAP) and Transformer attention mechanisms. The hybrid LSTM–Transformer model consistently achieved the best performance, with the lowest MSE and MAE and highest R2 across most runs. Statistical tests confirmed significant differences among models (Friedman p < 0.0001), with the hybrid outperforming CNN (p = 0.0286) and RNN (p < 0.0001). While LSTM and the hybrid showed comparable accuracy (p > 0.05). The hybrid model’s interpretability and performance make it a strong candidate for future real-world deployment; however, further work is needed to assess integration into educational systems, computational feasibility, and stakeholder acceptance. According to SHAP and attention studies, absenteeism, study time, and parental support were identified as the most influential predictors of performance, while gender and parents' level of education had little bearing. This research presents a balanced framework that advances both accuracy and interpretability in student performance modelling.

1 January 2023 in Journal of Information Technology Education: Innovations in Practice
Verified
Journal Article
Open access
PDF file
DDDelali Kwasi Dake
Delali Kwasi Dake
GBGodwin Kudjo Bada
Godwin Kudjo Bada
Aim/Purpose: The study focused on learner sentiments and experiences after using the Moodle assessment module and trained a machine learning classifier for future sentiment predictions. Background: Learner assessment is one of the standard methods instructors use to measure students’ performance and ascertain successful teaching objectives. In pedagogical design, assessment planning is vital in lesson content planning to the extent that curriculum designers and instructors primarily think like assessors. Assessment aids students in redefining their understanding of a subject and serves as the basis for more profound research in that particular subject. Positive results from an evaluation also motivate learners and provide employment directions to the students. Assessment results guide not just the students but also the instructor. Methodology: A modified methodology was used for carrying out the study. The revised methodology is divided into two major parts: the text-processing phase and the classification model phase. The text-processing phase consists of stages including cleaning, tokenization, and stop words removal, while the classification model phase consists of dataset training using a sentiment analyser, a polarity classification model and a prediction validation model. The text-processing phase of the referenced methodology did not utilise tokenization and stop words. In addition, the classification model did not include a sentiment analyser. Contribution: The reviewed literature reveals two major omissions: sentiment responses on using the Moodle for online assessment, particularly in developing countries with unstable internet connectivity, have not been investigated, and variations of the k-fold cross-validation technique in detecting overfitting and developing a reliable classifier have been largely neglected. In this study we built a Sentiment Analyser for Learner Emotion Management using the Moodle for assessment with data collected from a Ghanaian tertiary institution and developed a classification model for future sentiment predictions by evaluating the 10-fold and the 5-fold techniques on prediction accuracy. Findings: After training and testing, the RF algorithm emerged as the best classifier using the 5-fold cross-validation technique with an accuracy of 64.9%. Recommendations for Practitioners: Instead of a closed-ended questionnaire for learner feedback assessment, the open-ended mechanism should be utilised since learners can freely express their emotions devoid of restrictions. Recommendation for Researchers: Feature selection for sentiment analysis does not always improve the overall accuracy for the classification model. The traditional machine learning algorithms should always be compared to either the ensemble or the deep learning algorithms Impact on Society: Understanding learners’ emotions without restriction is important in the educational process. The pedagogical implementation of lessons and assessment should focus on machine learning integration Future Research: To compare ensemble and deep learning algorithms

1 January 2023 in Mathematical Problems in Engineering
Verified
Journal Article
Open access
PDF file
DDDelali Kwasi Dake
Delali Kwasi Dake
EGEsther Gyimah
Esther Gyimah
CBCharles Buabeng-Andoh
Charles Buabeng-Andoh
Counselling students remains a pre‐eminence for most tertiary institutions in Ghana to the extent that institutions now have counselling units that extend to the departmental level. This study used the K‐prototype machine learning algorithm to cluster students’ behaviour based on 28 relevant attributes and further proposed a classification model. The analysis of the experimental outcomes using the elbow method reveals the formation of three distinct clusters with decreasing intra‐cluster similarities and increasing inter‐cluster distances. The first cluster uniquely consists of active learners with three or more roommates, primarily in the first year. The second cluster with the highest membership consists mainly of second‐year students who exhibit passive classroom conduct and reside in a two‐occupancy hostel. The third cluster contains a mixture of third and final‐year students who are highly passive in class and live in a tenancy occupancy of two. After clustering, the K‐nearest neighbours, logistic regression, naïve Bayes (NB), and AdaBoost ensemble algorithms were implemented to create a model for future learner cluster prediction. Simulation results using the tenfold cross‐validation technique show that AdaBoost (NB) has the highest accuracy of 99.88% with an F‐measure score of 0.999 and receiver operating characteristic–area under the curve value of 1.00.