Hybrid of Kohonen Self Organising Maps in Rainfall Data Restruction
The issues on missing data particularly rainfall data have been the subject of discussions in many consultancy studies on water resources projects. The need for quality hydrological data especially rainfall data for planning, development and management of the country’s water resources has become increasingly important. This is due to the fact that water situation has changed from a relative abundance to a scarcity in some water-stressed parts of the world. Most hydrological services worldwide strive to have continuous rainfall records but only a few succeeded, especially in the developing countries. While every effort is made to ensure continuous rainfall data availability, data gaps in rainfall records still exist. This research is part of pro-active effort taken to explore techniques that can be adopted by data users in the treatment of missing rainfall records. The underlying principle of this study revolved around the application of combined Unsupervised Artificial Neural Network and Nearest Neighbour Imputation techniques. A Kohonen Self-Organizing feature Map (SOM) method is designed primarily for unsupervised learning. This proposed model, if properly applied could tremendously benefit data users.