Identification And Classification Of Highly Productive Lentil Varieties Affected By Disease Using Machine Learning Techniques

Abstract

Lentils Are One Of The Most Important Pulse Crops Produced In Ethiopia. Our Country Ranks On The Tenth Level In The World And First In Africa For The Volume Of Lentils Produced. The Most Significant Task In Agriculture Is Finding Improved Crops To Keep Food Security. The Agricultural Research Centers Focus On A Breeding Program To Improve Crops Production. Genetic Varietyidentification Between Genotypes Of Lentils Is Important For Selecting Genotypes For Breeding Programs. Accordingly, Agricultural Research Centers Doing Different Research On Breeding Programs To Improve Lentils Production, But Character Analysis Of Lentil Genotypes To Find Genetic Diversity And Lentils Yield Prediction Without Sowings Are The Active Problems. To Find The Solution To These Problems Machine Learning Techniques Were Proposed In This Study. The Data Used In This Research Was Collected From The Debre-Zeit Agricultural Research Center. Using The K-Means Clustering Algorithm, Centroid Initialized By K-Means++ Our Dataset Clustered Into Seven Different Groups Of Lentil Genotypes. Each Cluster Was Analyzed Using Traits Of Genotypes And Compared To Find Highly Productive Lentil Genotypes And Cluster VII Was Highly Productive. The Clustering Process Was Used For Labeling Datasets For The Classification Learning Algorithms. 10-Folds Cross-Validation Was Used To Evaluate The Regression Learning And Classification Learning Algorithms. The Classification Task Was Performed By Support Vector Machine(SVM), Random Forest(RF), Na??Ve Bayes(NB), And Decision Tree(DT) Models To Classify The Newly Inputted Genotypes. Each Algorithm Was Evaluated And Compared Based On Their Performance Evaluation Result, And The Support Vector Machine Has Better Accuracy Of 99.43%. The Yield Prediction Task Of Lentil Genotypes Affected By The Disease Was Performed By Random Forest(RF), Gradient Boosting(GB), And Multiple Linear Regression(MLR) Models.Those Models Were Implemented Without And With UFS And RFE Feature Selections. The Random Forest Has Better Performance With MAE=17.51, MSE=1089.18, R-Squared=0.96. The Recommendation Of Lentil Genotypes For The Breeding Process Task Was Applied By A Collaborative-Based Filtering Algorithm. This Study Reveals That Machine Learning Algorithms Are Important In The Agricultural Sector, To Cluster, Classify, Yield Prediction, And Recommendation Of Lentil Genotypes. We Hope This Study Will Be Helpful For More Findings Concerning The Identification And Classification Of Lentil Varieties Using Machine Learning.