Forecasting Bush Mango (Irvingia Gabonesis) Fruit Yields Using Machine Learning Models

ABSTRACT

In this study a comparative analysis of three machine learning models, Decision Tree (DT), Support Vector Machine, (SVM), and Neural Network (NN) were compared to predict and forecast the number of fruits per tree or an accession of Irvingia Gabonesis at every season for farmers to plan adequately for future harvest and to forestall wastages.

Availability of I.Gabonesis is being constrained by lack of intensive cultivation, absence of plantation, ignorance of the use of its by-products, poor marketing and perishability of its mesocarp and kernel (seed). The success of the research work is to assist farmers to know how to plan for the fruit tree species in terms of harvest and economic benefits

Bigml API and Python language were used. The dataset was cleansed, loaded into the model, feature extractions/engineering were carried out. The dataset was split into ratio 80:20 for both training and testing respectively. The result gotten were subjected to three performance metrics: Mean Absolute Error (MAE), Mean Square Error (MSE) and R Square to achieve an optimum result among the three selected models. The models’ results were as follows: DT, SVM, and NN 0.70, -0.422, and -I.015 respectively with DT having the best result of 70%. In addition to this, Visualizations of nofruitplanted against plantno using line graphs were performed.

Keywords: Irvingia gabonesis, machine learning models, fruition, forest, API, Python.