A Dynamic Model For Prediction Of Food Insecurity

ABSTRACT

The world population experiencing food insecurity ranges from 0.76 to 1 billion people [7]

[27] [30] [77]. This is approximately a seventh of the world population. In the report released

by Food and Agricultural Organization (FAO) [30], 843 million people were categorized as

victims of chronic hunger. It was further established that approximately 805 million people

were undernourished. Food insecurity is therefore a formidable challenge affecting the world.

Several factors are said to exacerbate this challenge. According to FAO [26], expanding

populations and growth in biofuel practices are exerting high demands on the available food

systems. Soil exhaustion, climate changes and environment degradation are on the other

hand negatively affecting agricultural yields thus further fueling food insecurity challenge.

Poverty has also been cited as yet another catalysit for this challenge: low income populations

face financial constrains in acquiring sufficient food for dietary intake [60]. Factors such as

these are making it difficulty to realise the first Millennium Development Goal (MDG1) of

reducing the victims of hunger world wide [26].

The importance of predicting or monitoring food insecurity has been highlighted by

various studies [16] [54]. If this disaster is forecasted early, stakeholders can take necessary

steps to prevent it or control its impact. This has proved successful in some parts of the world.

For instance, according to the United States Department of Agriculture (USDA), reliable

monitoring of food insecurity contributes to the effective operation of Federal programs, food

assistance programs, and other government initiatives aimed at reducing food insecurity [71].

Machine Learning (ML) is gaining momentum in the arena of modeling food dynamics.

It is an emerging discipline in computational science whose application is also useful in other

areas such as disease detection/diagnosis, web search, spam detection, credit scoring, fraud

detection, stock trading and drug design [23] [67]. In relation to food security, machine

learning techniques are well suited for prediction of risks like famine since they can enhance

2

classification accuracy [54].

The Role of Machine Learning in predicting food insecurity has been exemplified. Contributions

to knowledge gaps in areas such as performance improvement, teasing predictions

using various data sets, fine tuning parameter selection and exploring the application of Machine

Learning have been instrumental in fostering the science of predicting food insecurity.

These studies have however given little attention if any on re-usability of machine learning

models. In this study we aim at developing a Dynamic Model (DM) to predict food insecurity.

The model will use other existing models as a technique to enhance prediction in an

environment of limited data sets.