Developing Knowledge Based System for Cereal Crop Diagnosis and Treatment: The Case of Kulumsa Agriculture Research Center

No Thumbnail Available



Journal Title

Journal ISSN

Volume Title


Addis Ababa University


The aim of this research is to extract hidden, novel and potentially useful knowledge concerning corruption from the data taken from existing database of the FEACCE with the help of data mining techniques and tools selected. Persons who are working in any activity can commit corruption knowingly or unknowingly but we don’t know what persons with what personal characteristics are vulnerable to corruption. Currently the FEACCE staffs who are working in investigation of corruption attempt to understand few relationship between corruption category and the characteristics of offenders. This can be achieved by utilizing data mining techniques efficiently, effectively and accurately than those staffs who are working in statistics department using traditional and simple statistical methods to analyze corruption data. To attain the objective of the study the researcher has used CRISP-DM methodology that consist six steps and the steps can be used iteratively until the required results are achieved. From the steps, data preprocessing needs to be given the higher priority because the data bases of FEACCE was inconsistent and incomplete and this needs to be cleaned before it is given as an input for the data mining tool. This step needs more time and effort than the remaining steps. Thus results from data mining techniques that were relied on the structured approach were useful for the FEACCE to attain its objectives. This study applied three different data mining techniques and came up with different models along with evaluations. The models can be used by the FEACCE and other bodies that are combating corruption in the country to predict the future events and classify corruption offenders through both predictive and descriptive modeling techniques. Some of rules generated from association rule mining were not that much interesting because some of the attribute values in the data base were insufficient and some are many. Approximately rules generated from association rule mining and rules developed in classification model were the same as they both are predictive models. The classification model that uses the output of clustering model as an input has best performance than the direct classification of the data set. Application of data mining in corrupt activity data is introduced in this research.



Corrupt Activity Data to Support Combating Corruption