An indigenous knowledge-based weather prediction platform using fuzzy logic for fishing communities along Lake Victoria

The indigenous fishing community along Lake Victoria relies on traditional knowledge methods for weather forecasting. This study utilized Mamdani-type fuzzy logic to preserve the continuity of indigenous knowledge while also seeking ways to enhance its effectiveness. The research began by examining 17 indigenous knowledge indicators and later employed Random Forest regression and proportional Piling methods, identifying Nimbus Clouds, Wind Circulation, Moon Phases, and Body Feel indicators as the main indicators for daily forecasting. Data collected during this period represented both the dry and rainy seasons, which were the primary seasons of interest. The daily predictions generated by the system were compared with actual weather conditions, yielding an accuracy rate of 63.5% for the fuzzy-logic-based indigenous knowledge weather predictions after thorough redundancy and inconsistency checks on the fuzzy rules. However, the method had drawbacks, including insufficient training datasets and the qualitative nature of IK indicators, which relied on standard scientific scales to generate fuzzy-logic rules. Despite the challenges, fuzzy logic-based indigenous knowledge enhanced the dissemination of weather predictions in a local language and in an accessible format, thereby improving the community's access to weather and climate interventions and ability to adapt to climate risks. The method offers a practical and economical an ICT-based delivery that enhances the functionality, storage, and application of traditional knowledge for future use. This research, therefore, establishes a strong foundation for further exploration of integrating indigenous knowledge with modern climate science methodologies to improve outcomes.