FACULTY OF COMPUTING,

The rapid digital transformation across global industries has intensified the demand for modern technical skills, exposing significant skill gaps within African labour markets. Traditional labour forecasting methods, which rely on surveys and static reports, struggle to capture evolving workforce trends and often fail to provide timely, data-driven insights. This limitation motivates the adoption of machine learning-based analytical models capable of identifying emerging skill patterns from historical job market data. This study aims to develop and evaluate a predictive system for forecasting skill demand using structured datasets collected from African job portals. Specifically, the research applies Linear Regression as the core forecasting model to estimate future demand trends for key digital skills. The methodology involves a complete data preprocessing pipeline consisting of cleaning, normalization, and restructuring job-skill frequency data into model-readable formats. A curated dataset covering multiple skills across four years was used, and the system was implemented using Python and Streamlit for interactive visualization. Model performance was assessed using accuracy, trend-direction consistency, and graphical evaluation metrics derived from observed versus predicted values. Results show that the Linear Regression model accurately captured general growth patterns for high-demand digital skills such as Python and Data Analysis, achieving strong alignment between historical trajectories and forecasted values. The deployed system demonstrated stability, fast response time, and ease of use, enabling real-time skill trend visualization and one-step-ahead forecasting. The study confirms the potential of machine learning approaches for supporting labour market analysis in Africa. However, limitations include restricted dataset size, reliance on numerical trend features, and absence of deep learning models such as LSTM, which may better capture complex temporal dependencies. Future research should incorporate larger datasets, integrate natural language processing for extracting skills from job descriptions, and extend the forecasting engine to more advanced time-series models for improved prediction accuracy and adaptability.

This project focuses on the design and implementation of a lightweight honeypot system aimed at detecting and logging unauthorized access attempts on a local network. With cyberattacks
becoming increasingly common and sophisticated, many small organizations lack the tools to properly monitor their networks. Honeypots offer a simple yet effective way to study intrusion behaviour by acting as decoy systems that attract potential attackers. The system developed in this project was built using Python due to its flexibility and strong support for network programming through libraries such as socket, datetime, and logging. It works by listening on a specific network port, recording each incoming connection, and logging details such as the intruder’s IP address, timestamp, and input commands into a text file named honey.txt. Using the prototyping methodology, the system was built and tested within a controlled local
environment using Telnet and Nmap to simulate intrusion attempts. The results confirmed that the honeypot could efficiently detect connections, capture relevant data, and operate smoothly with minimal resource use. Overall, this project shows that lightweight honeypots can provide valuable insights into network security and help raise awareness of potential threats, especially for small organizations and students. Future improvements could include multi-port monitoring, real-time alert features, and database-based log management to enhance system efficiency and scalability.

Sales order accounting system development is essential for modern retail businesses seeking to eliminate the inefficiencies of manual sales processing, inventory tracking, and financial reporting. This project presents the design and implementation of a fully automated, client-side sales order accounting system built with HTML5, CSS3, vanilla JavaScript, and localStorage. The system integrates seamless sales order creation with unique identifiers and real-time input validation, dynamic inventory management featuring instant stock updates and proactive low-stock alerts, and comprehensive financial reporting modules that generate daily sales summaries, profit-and-loss statements, and exportable transaction logs. It also supports basic marketing functions by enabling secure export of customer purchase records for personalized email or SMS promotions. The application runs entirely offline, supports cross-session data persistence, and features a clean, intuitive, mobile-responsive interface, making it suitable for small businesses with limited technical infrastructure. Although Shoprite is referenced as an example of a large retailer already using computerized systems, this solution is specifically designed to provide small businesses with an affordable, scalable blueprint for digital transformation. Rigorous testing showed 98% accuracy in transaction processing, a 70% reduction in order-processing time, and a 4.7/5 satisfaction rating from test users, demonstrating its reliability, efficiency, and practical value in real retail environments.