Browsing by Author "Yusuf, I."

Now showing 1 - 3 of 3
  • Thumbnail Image
    Item
    Development of a Bank Customers’ Credit Rating System Using Neural Networks Algorithms
    (Faculty of Science, Gombe State University, Nigeria - Bima Journal of Science and Technology, 9(1A), 233 – 244, 2025) Abdulsalam, S.O.; Yusuf, I.; Awotunde, S.O.; Edafeajiroke, M.F.
    Credit rating systems are vital for financial institutions to evaluate creditworthiness, manage risk, and minimize losses from defaults. Traditional methods, relying on historical financial data, lack adaptability and exclude borrowers with limited credit histories. Machine learning (ML) techniques, such as Backpropagation Neural Networks (BPNN) and Feedforward Propagation Neural Networks (FPNN), offer improved predictive accuracy. This study developed an advanced credit rating system using BPNN and FPNN, trained on Nigerian bank data incorporating financial and non-financial indicators. BPNN’s iterative weight-adjustment capability through backpropagation enhanced performance, while FPNN served as a simpler baseline. The methodology included data collection, preprocessing, model training, and evaluation using accuracy, precision, recall, and F1-score. Results demonstrated BPNN’s superiority, achieving 82.5% accuracy compared to FPNN’s 81.2%, alongside higher precision, recall, and F1 scores. The study underscores BPNN’s effectiveness in refining credit risk assessment, making it a more reliable tool for Nigerian banks. By leveraging ML, financial institutions can enhance decision-making, reduce risks, and expand credit access to underserved populations. Keywords: Credit Rating System, Back Propagation Neural Networks (BPNN), Feed-Forward Propagation Neural Network (FPNN), Development of Banks' Customers Credit Rating System, Machine Learning in Banking system
  • Thumbnail Image
    Item
    Evaluating the Effectiveness of Social Media Marketing Strategies on the Performance of Small Businesses in Nigeria
    (Journal of Family Business and Management Studies, Business School, University of the Thai Chamber of Commerce, Bangkok, Thailand, 2026-01-28) Salau, A. A.; Yusuf, I.; Adebayo Adeyemi Abdulwasiu; Akinwumi, D. S.
    Today, small businesses small businesses are faced with plethora of issues that stall their performance and marketing processes because there is absence of adequate resources to invest in mainstream marketing programs which birth the inability to fully achieve the identified objectives The study examined the effectiveness of social media marketing strategies on the performance of small businesses in Nigeria. Also, four research objectives and hypotheses were developed. The study utilized a cross-sectional survey design. The population of the study was 3,005 selected registered small business owners, and a sample of 353 was arrived at using Taro Yamane’s formula. The questionnaire was developed in a structured and closed-ended form, administered by Google Forms. Out of the 353 questionnaires, 350 were filled, giving a response rate of 99.1%. A simple random sampling technique was adopted to ensure that each small business within the population had an equal chance of being selected, thereby minimizing bias and improving the representativeness and generalizability of the findings. A Likert scale was adopted ranging from 1 (Strongly Disagree) to 5 (Strongly Agree). To guarantee the quality of the study’s instrument, face validity was checked, experts reviewed content accuracy and inter-observer reliability was measured with a Cronbach’s Alpha coefficient value of 0.82 for performance, 0.78 for influencer marketing, 0.80 for user generated content and 0.84 for paid advertising which revealed the acceptable level of 0.70. The collected data was analyzed using descriptive and inferential statistics. Hierarchical regression analysis was used to test the hypotheses using SPSS version 27. The findings of the study revealed that all four strategies had a positive and significant effect on the performance of small businesses in Nigeria. Among all the social media marketing strategies, user-generated content (B = 0.4789, p = 0.003) and community engagement (B = 0.4023, p = 0.013) proving that being open and interacting leads to more achievements. Although influencer marketing (B = 0.5125, p = 0.002) and paid advertising (B = 0.3265, p = 0.039) had a positive effect, but to a smaller extent. The result revealed that all the four variables exhibited positive and statistically significant relationships with business performance: Influencer Marketing (B = 0.5125, p = 0.002), Paid Advertising (B = 0.3265, p = 0.039), Community Engagement (B = 0.4023, p = 0.013), and User-Generated Content (B = 0.4789, p = 0.003). regression model was statistically significant indicating that the four predictors jointly explained 20% of the variance in small business performance. Regression model was statistically significant indicating that the four predictors jointly explained 20% of the variance in small business performance Diagnostic tests confirmed that the model satisfied assumptions of normality, homoscedasticity, multicollinearity (VIF = 1.83–2.41), and residual independence (Durbin–Watson = 1.89), validating the robustness of the results. Based on the result of the study, it was recommended that small businesses should focus on having customers post content and interact regularly to earn their trust and succeed in the digital world for a longer time.
  • Thumbnail Image
    Item
    Predicting Customers’ Churn in a Bank Using Firefly Feature Selection and RBF-Support Vector Machine Classification Algorithms
    (Faculty of Information and Communication Technology, Kwara State University, Malete, Nigeria - KWASU Journal of Information, Communication and Technology, 2(1), 17 – 25, 2023) Abdulsalam, S.O.; Yusuf, I.; Ogunwuyi, M.R.; Edafeajiroke, M.F.
    Customer churn prediction helps businesses identify factors contributing to customer attrition, allowing them to take preventive actions. This project develops a churn prediction model for a bank using Firefly Feature Selection and RBF-Support Vector Machine (SVM) Classification Algorithms. The dataset consists of 900 instances and 10 features, representing customer churn rates. Firefly Feature Selection was used to find the most pertinent features after preprocessing techniques like Label Encoding and Oversampling were used. After that, the RBF-SVM model was used to make predictions. F1-score, recall, accuracy, and precision were used to assess the model’s performance. The strong predictive capability was demonstrated by the model’s 84% accuracy, 84% precision, 100% recall, and 91% F1 score. This methodology assists companies in identifying consumers who are at risk and in taking preventative action to lower churn rates. Future research could focus on improving feature selection techniques, model robustness, and applying the model to other industries for comparison. Keywords: Customer churn prediction, Firefly Feature Selection, RBF-Support Vector Machine (SVM), Classification algorithms, Model evaluation