MSc. CIS Theses and Dissertations (2023)

Permanent URI for this collection

http://hdl.handle.net/11071/13544

Browse

Recent Submissions

Now showing 1 - 2 of 2
  • Thumbnail Image
    Item
    Harnessing tacit knowledge to improve employee performance using AI Voice detection - a case of Kenya Railways Corporation
    (Strathmore University, 2023) Maina, D. A.
    Harnessing knowledge in organizations is important in improving employee performance. Explicit knowledge is widely shared because of its descriptive nature and easy documentation. Tacit knowledge is under-utilized due to its intangible nature. It is knowledge based on experience embedded in a person. Tacit knowledge is gained from the continuous practice of organizational tasks, which helps build valuable experience, intuition, innovation, and better ways of handling situations. Experienced employees in an organization have more tacit knowledge compared to younger employees. When faced with a challenging situation at the workplace, younger employees need to consult experienced employees on the best way to tackle; if there is no one to consult they would have to try out their way or make mistakes and learn from them. When these older employees leave the organization they leave with a wealth of tacit knowledge embedded in them. Due to the lack of an efficient channel to share and store tacit knowledge, Kenya Railways loses loads of information that could help smoothen business processes save time and money, and improve the performance of its employees. Transfer of Tacit knowledge is crucial to Kenya Railways Nairobi Central Workshop because of the unique nature of its operations. To fill these gaps, this study explored the use of AI Voice detection to harness tacit knowledge. AI voice detection system was used to capture tacit knowledge in audio form and stored it in the knowledge base. Upon a user’s request, the system base is queried to give the required feedback. The development of the AI voice detection system adopted Agile Software Development Methodology. This methodology is an iterative and incremental approach to software development. The data collected targeted engineers and technicians in the Nairobi Central Workshop working on the repair of Locomotives and DMU. The data included sources of tacit knowledge, challenges in sharing it, areas that require the tacit knowledge, and users’ functional requirements of the bot. From the challenges identified tacit knowledge was gathered and fed into the bot. This information was used to constantly train the model to increase its efficiency in delivering tacit knowledge to users without human intervention. The data collected was classified into three broad categories: Training set, Test set, and Validation set. The training used supervised learning where the bot learned from labeled datasets.
  • Thumbnail Image
    Item
    Retailer stock levels optimization tool
    (Strathmore University, 2023) Ndana, J.
    The purpose of this research was to design a statistical model that allows a retailer to optimize stock levels based on stock related parameters such as demand, lack of stock, stock replenishment lead time, service level, maintenance cost, and costs of replenishing stock. The study adopted applied research to solve the business challenge on optimization of stock. The study utilized secondary historical data relating to stock obtained from a supermarket in developing the optimization model. Additionally, the study applied prototyping methodology to design, develop and test the prototype. The web application was developed using HTML, JavaScript and Java Server Pages in Netbeans IDE. The server applications were developed based on Java Programming language. Apache Kafka was used for ingestion. Spark was used for data streaming while YugabyteDB was used for storage. The model, based on parameter values of a high moving product, yielded a service level at 95.2%. This was a positive indicator that the retailer would not hit a stock-out during the subsequent replenishment cycle for this product. On a probability scale of 0.01 to 0.99, the probability of running out of stock was 0.048. The model yielded an optimum order quantity of 15 units, against an average of 17 units on supplies made. Moreover, the model computed an optimum safety stock level within the 10-20% range of 14 units, which allows the retailer to cater for varying vendor delivery periods, as well as meet the changing consumer demands. Based on these values, the model computed an optimum stock level of 21 units, which allows the retailer to only reorder when the cycle stock, computed at 7 units, nears depletion. Similarly, the retailer can further inform decision making in the reorder placements based on the computed average lead time, such that the delivery is made every 6 days. The model was further tested on another high moving product, yielding a service level of 96.5%, implying that approximately 96% of the periods the model is able to cover for the customer demand for the given product.