ANAYO, NNENNA GRACE (2026) ENHANCED FARM YIELD PREDICTION SYSTEM. Other thesis, GODFREY OKOYE UNIVERSITY, ENUGU STATE.
|
Text
Anayou Nnenna Grace Final Doc. (7).pdf Download (2MB) | Preview |
Abstract
Despite agriculture being the mainstay of the economy in Nigeria, small scale farmers in the country still depend on subjective and manual methods for making predictions on their yield, hence ineffective plans leading to food insecurity. In this paper, an enhanced system for predicting the farm yield of Nigeria’s native crops was designed using two catboost algorithms; CatBoost classifier that helps in recommending crops and a CatBoost regressor that makes yield prediction for those crops, which were all hosted on a website developed using React 19 and Flask framework. A dataset containing 20,000 records from Nigeria’s agricultural sector that had data of nine crop types, six geopolitical zones, 36 states and the Federal Capital Territory, and five agro-ecological zones was downloaded from Kaggle and preprocessed using a six step pipeline comprising of column dropping, stratified median imputation, and creation of seven new columns. The CatBoost Classifier gave a classification accuracy of 89.5 percent in a test set of 4,000 records, giving a Macro F1-Score of 0.8132 and Weighted Precision of 90.0 percent. The CatBoost Regressor gave an R2 of 98.48 percent, RMSE of 435.9 kg/ha, MAE of 308.8 kg/ha, and MAPE of 11.36 percent. Feature Importance Analysis validated that Humidity and Potassium levels in soil are the key feature for crop recommendation and Crop Type explains 60 percent variability in the yield data. This system is the only one mentioned in the relevant literature which has features like full deployment on web, covergae for nine crops in Nigeria, dual task catboost and agronomy recommendations all together in one platform.
| Item Type: | Thesis (Other) |
|---|---|
| Subjects: | Q Science > Q Science (General) |
| Divisions: | Faculty of Engineering, Science and Mathematics > School of Electronics and Computer Science |
| Depositing User: | Nnenna Ayo |
| Date Deposited: | 22 Jul 2026 10:04 |
| Last Modified: | 22 Jul 2026 10:04 |
| URI: | http://eprints.gouni.edu.ng/id/eprint/5973 |
Actions (login required)
 |
View Item |
