Artificial Neural Networks for Predicting Coffee Rust Case Study

After looking at the nuts and bolts of natural language processing in my last blog, today I want to look at how artificial neural networks (ANNs) can be used to predict coffee rust. This is a sneak peek at the  research that I will present at the 2019 SciPy Conference.

But you ask, what is coffee rust and why should I care about it? Coffee rust infestation leads to production losses of over $1 billion annually worldwide and you will probably notice the effects of the disease on the price of your morning joe. Coffee rust is the most deadly disease that attacks the coffee plant and is caused by the Hemileia vastatrix fungus at temperatures between 15-28° C.

Before we can talk about how artificial neural networks were used to predict coffee rust, it’s important to understand what an artificial neural network is. Jahnavi Mahanta has a great definition of artificial neural networks as using “ the processing of the brain as a basis to develop algorithms that can be used to model complex patterns and prediction problems.” An ANN is a simple representation of what is happening in the brain. Neurons in the brain process information using billions of neurons in the cells. The information is processed and then is converted to some sort of output. In a similar way in an ANN, information comes into an input layer, is processed in one or more “hidden layers” where it learns and then produces some type of output. There are much more technical explanations of ANNs out there but in order to stay true to the intended non-technical audience of this blog, I will end my explanation here.

Let’s get to the coffee rust example now. Like any data scientist, I started with a hypothesis that coffee rust could be predicted with data. I used four variables/ features as inputs to the ANN. My target output is the amount of coffee rust in a coffee-growing region on a weekly basis. After collecting the data and exploring it a bit, I started with basic machine learning models. Since those did not perform well and because of the dataset’s non-linearity, I moved on to ANNs since one of the advantages of an ANN is that it can derive meaning from imprecise data such as my coffee rust dataset. The ANN performed over 100 times better than the other machine learning models and indeed can be used to predict coffee rust.

Thanks for reading. In my next blog, I’ll talk about how artificial neural networks can be used for image classification.

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.