تطوير نظام الكتروني للتنبؤ بمعدلات هطول الامطار باستخدام النماذج التجميعية في تنقيب البيانات دراسة حالة (جمهورية السودان)

Abstract

Abstract In the twenty -first century, with the development of information tech- nology, an evolutionary boom occurred in the form and size of the information, then developed tools raised to deal with more precisely and more performance. In the past, statistical methods are used to support decision-making and recent- ly developed to data mining to keep pace evolution, and it includes several technologies, including prediction technology. Research deals with several problems, they are, the current systems aren’t accurate and not stable when building models with mono algorithms due to the presence of weaknesses, not using other features with a great impact in developing prediction models, diffi- culty of accurate rain fall rate value prediction, and high economic costs and poor funding for these current systems. Scientific methodologies include the analytical, descriptive, historical, inductive, applied scientific, and experi- mental methodologies, with the aim of developing electronic System using de- veloped accurate and stable an ensemble model to predict the categories of rainfall rates by building a model for selecting the best features that achieve the highest accuracy in prediction, and to predict the categories of rainfall rates during 24 hours, it uses five Algorithms for prediction that were chosen from making experimenting on ten prediction algorithms according to the highest accuracy and adequate training using large data sizes from NASA with 216.972 records and 35 features in the period from January 2000 AD to December 2021 AD for 27 station of the Republic of Sudan, then combine the ensemble model into an electronic system by developing it using the object oriented methodolo- gy, and conducting experiments on it. The most important results of research are, developing accurate and stable an ensemble model which is built from se- lecting the best features by using forward selection algorithm in terms of accu- racy and they are twelve best features, an ensemble model is developed to pre- dict standard universal categories in millimeters for rainfall rates, with evaluat- ing the ensemble model results accuracy rate 78%, precision rate 80%, Recall rate 78% , and F-score 79%, according to the experiments, the stable accurate ensemble model in an electronic system predicted rain fall rates categories with high matching percentage by comparing them with the real categories of rain fall rates in most Data parts according to random sample.