Prediction of Maize Yield Using Machine Learning

Author

Zvarevashe, Zivanayi

Title

Prediction of Maize Yield Using Machine Learning

Abstract

The agricultural sector is a critical component of the Zimbabwean economy and the future of humanity. The aim of this research was to develop a machine learning model for predicting maize yield and to evaluate its accuracy and performance. The model is based on a range of factors that impact maize yield, including historical yield data, weather conditions, soil characteristics, and planting practices. The study had three (3) research objectives, the first one was to analyze and identify variables necessary for predicting yields using machine learning algorithms, the second objective was to develop machine learning models for predicting maize yield based on relevant features such as weather patterns, soil type, and fertilizer application. The last and third objectives was to evaluate the effectiveness of machine learning algorithms in predicting maize yield. The author used waterfall software development model to create the machine learning model using the Linear Regression algorithm. The model was trained using a dataset with approx 1000 rows with necessary variables to predict maize yield. The author used various data-preprocessing techniques and managed to fit the algorithm in the dataset. Flask web framework was used as a web client framework for displaying the system on the browser. The author used the confusion matrix to evaluate the performance of the model. The model was able to predict the maize yield with high accuracy. The results showed that the model was able to predict high-yield and low-yield samples with an accuracy of 94%, a precision of 95.8%, an F1 score of 93%, and a recall of 92%. The model had a misclassification error rate of 6%, which indicated that it incorrectly classified 6% of the low-yield samples as high-yield. Overall, these findings suggest that the machine learning model and web app have the potential to be used as a reliable tool for maize yield prediction. These findings suggest that the logistic regression model is a useful tool for predicting maize yield, and could be employed by farmers and agricultural policymakers to make informed decisions regarding crop management and resource allocation.

Date

2023

Publisher

BUSE

Keywords

Prediction

Maize Yield

Machine Learning

Supervisor

Mr Chaka