Genome Based phenotype Identification Model For Chickpea using Machine Learning Approach

Ethiopia is the leading producer, consumer, and seller of chickpea in Africa and one of the
top ten most important producers of chickpea in the world. Chickpea plays an important
role in poor people’s diet and human nutrition in Ethiopia and the arid and semiarid regions
of the world. Genome enabled prediction technologies trying to transform the identification
of crops phenotypes and not only change but upgrading the existing prediction and
identification paradigm towards the next generation of agricultural breeding mechanism.
Current state of the identification of chickpea phenotype in Ethiopia still sticks to a manual
system. Domain experts tried to recognize every chickpea genome and phenotypes, the
way and efficiency of identifying each chickpea varieties mainly depend on the skills and
experience of experts in the domain area. Attempts are made to design phenotype
identification model to facilitate the selection, identification and classification of different
crops phenotypes but Most of those researches were done outside Ethiopia; for local and
emerging varieties, new paradigms should be researched and even the accuracy of an
existing algorithm should be verified. As to the researcher knowledge no studies are
conducted to find a way to design such model in the area of Ethiopian existing and
emerging chickpea varieties therefore is a need to design identification model that assists
the selection and identification mechanisms of chickpea.
The main aim of this study is to design phenotype identification model using machine
learning algorithm that identifies chickpea phenotypes. To build the identification model
of chickpea phenotype artificial neural network, support vector machines and decision tree
were used. For evaluating the performance of the developed model confusion matrix with
accuracy, recall, and precision were used. A total of 8303 records 80% for training and
20% for testing with 9 features were used in order to identify chickpea phenotype. Data
collection, data preparation, feature selection, data preprocessing and data transformation
were done in order to prepare the dataset for experiments. The evaluation of best
performing algorithms is compared according to accuracy, sensitivity, specificity and precision. Based on performance evaluator the best algorithms are found to be decision tree
and achieves 97.5% accuracy.
Finally, the main outputs of this study for domain experts and research community is design
chickpea phenotype identification model. Another important output is collecting and
preparing chickpea genotype dataset based on the feedback from domain experts of Debre
Zeit agriculture research center in order to help other researchers in conducting related
studies to handle data problems.