A Novel Multi-class Classification of Obesity Level using Artificial Neural Network Machine Learning Model

Obesity which is the excessive accumulation of fat impairs an individual health status and thus contributes to numerous chronic diseases, including cancers, diabetes, metabolic syndrome cardiovascular diseases among others. To efficiently predict obesity using the obesity dataset needs a good computational technique such as machine learning techniques that have the power to eradicate data inconsistencies and detect patterns and behaviors of obesity occurrence. This paper aims at predicting the occurance of obesity using the obesity dataset obtained from the University of California Irvine (UCI) machine learning dataset repository. The dataseet was preprocessed by Identification and handling of missing values, Encoding the categorical variables as well as Scaling the dataset before splitting it for training and testing the artificial neural network. The Artificial Neural Networ model was built in python programming language using the Jupiter NoteBook as the programming environment along side with other python third-party libraries such as Numpy, Pandas, Sklearn, MatplotLib and Keras. The Artificial Neural Network model performce very well in the multiclass classification of obesity with prediction accuracy of 97%.