In this article, the JENGA School Cohort 3 group, who are currently working on various data science capstone projects, share their take on how artificial intelligence can help to solve some of Africa’s challenges.
Artificial Intelligence in Health Care
Health care in Africa suffers from a multitude of challenges, ranging from misdiagnosis to lack of skilled personnel and even lack of specialized equipment. Unknown to many, artificial intelligence (AI) can actually help to solve some of these issues and improve the quality of health care on the continent.
For example, the use of artificial intelligence can be intensified to aid in cancer research in Africa. Data from imaging tests can be used to develop algorithms that can accurately find cancer and avoid misdiagnosis, identify the stage and aggressiveness of a tumor and inform doctors whether treatment is working or not.
~ Dr. Lawrence Nderu
Using Data Science to Combat Air Pollution
On a typical day in Nairobi’s hustle and bustle, you are bound to breathe in polluted air. If it is not from the noisy rowdy matatus, then it is from a nearby factory or the wind blowing a cloud of dust your way. Breathing this kind of air is not only uncomfortable but also unhealthy. The WHO reports 7 million deaths annually around the globe caused by air pollution. It doesn’t end there, air pollution is one of the causes of humanity’s greatest challenge in the 21st Century – Climate Change.
Knowing how deep air pollution cuts, what can data scientists do to combat these effects? One of the solutions is to track the pollution. My project seeks to collect and sensor data across Nairobi, Kenya, visualize and analyze it to understand the trends in a particular pollutant – PM2.5. Insights from this project could help policymakers, lawmakers and active citizens understand the air they breathe and take necessary action towards ensuring cleaner air in Nairobi.
~ Sifa Kinoti
Can Data Science eradicate Insurance Fraud?
Motor Insurance is a legal requirement mandated by Chapter 405 of the Kenya Traffic Act. Therefore, every car, driven on Kenyan roads should have insurance and display it. This got me thinking about areas of motor insurance where data science can be applied. I landed on one of the largest and most well-known problems that insurance companies face; fraud.
Processing fraudulent claims can be highly expensive for insurance companies which end up incurring huge losses. These companies in turn inflate their premiums to recover what is lost. Man eat man society, right
However, the best way to combat this is to take advantage of the largest asset that insurers have, which is data. This can be done by using various machine learning and deep learning models for fraud prediction.
These models are then compared to see which one is more effective, using different performance metrics. The aim is to provide a model with a great ability to identify in-depth patterns in data that are normally invisible or difficult to identify using other methods. These patterns can then be used to detect and predict fraudulent claims.
~ Joy Grace Ngugi
Problem-Solving Africa’s Water Crisis Using Data Science
The phrase ‘only in Kenya’ is one that we have all heard or used as Kenyans. It is used to point out some of the absurdities and ironies that exist in society. For example, it is only in Kenya where flooding seasons will always be followed by severe water shortages. Absurd, right?
I am not whether this scenario only plays out in Kenya but it is a fact that water shortages are a problem both in the developed and developing worlds. As extreme weather conditions worsen and populations continue to grow, so is water scarcity expected to grow.
How can data science help?
- Data science can help make better use of existing water resources.
Utility companies, manufacturing, and agricultural industries are some of the biggest water consumers. In fact, it is said that they consume twice as much water as individual households. These industries can use analytic tools to improve efficiency and reduce wasteful water use, therefore lessening their impact on the water crisis.
- Data science can provide real-time systems for monitoring water availability and quality so that action can be taken at the right time.
Data from water sensors or water samples can be monitored to detect changes in things like PH levels, temperature, oxygen, and salinity. If trends suggest that water may soon run out or become unusable, organizations and governments can take appropriate action immediately.
Using predictive machine learning models to forecast groundwater levels and stream flow levels in areas that are at risk of facing water shortage will allow for preventive and mitigative measures to be taken in good time.
Flood forecasting systems and rainfall run-off estimation models can also be helpful to governments and organizations, not only in preventing potential harm but also in preparing adequate water harvesting and storage infrastructure.
In a nutshell, with data science, we can use existing information and data to better understand the problems associated with the water crisis as well as formulate potential solutions. This will move us a step closer to a solution to the water crisis in Kenya, and Africa as a whole.
~ Felista Mogire
Using image segmentation from satellite data for slum detection
Today whenever you visit an unfamiliar area, you will most likely turn on the GPS location feature on your phone, and follow the directions to your desired destination, right? But do we take time to understand how these functions work?
In the last decade, we’ve seen numerous satellite launches with more than 1000 satellites launched in 2020 alone. Many of these satellites are used by Google in Google Maps to provide location information, which is frequently used on different devices. Most of these satellites provide readily available data with a wide range of applications, which are not limited to big corporations. Anyone with a basic understanding of image processing can use satellite data to extract meaningful information from images.
In data science, we strive to use available data to create solutions to real-world problems. With satellite images, the biggest challenge we have is not the lack of data, but extracting meaningful insights from the data.
Out of interest and my own research I settled on the slum detection and mapping project. Image segmentation from satellite images is an area that is little explored, which is why I decided to take on the challenge.
This project aims to detect urban areas with informal characteristics from satellite images. This data will specifically be useful for urban planning and allocation of resources to ensure that all urban dwellers have equal access to facilities and services.
~ Gacheri Nturibi
Anomalies and anomaly detection using artificial intelligence
Outlier or anomaly detection is a key step in data mining that entails identifying rare and peculiar patterns in data. Anomalies can occur as a result of a change in system behavior, human error, malicious activities, instrument error, and fraudulent activities.
Detecting such events offers significant information that can trigger critical actions in various application domains. While some domain fields recommend that such events be discarded, data science enthusiasts envision that such events could be signaling something, which if ignored can cause catastrophes.
In the past, simple statistical methods such as standard deviation and variance were used for anomaly detection. Today, things have changed. The internet has dramatically evolved and the number of devices connected to the internet is growing exponentially. These devices generate data at an ever-increasing speed, in large volumes, velocity, veracity, and variety. This is commonly known as “big data”.
As a result, detecting anomalies in such data requires efficient and improved anomaly detection techniques, which is where artificial intelligence comes in. Anomaly detection algorithms in AI can be used in various sectors such as detecting fraud in financial transactions, cyber-attacks within a computer network, or even faulty machines and processes in the manufacturing industry.
~ Daisy Nyang’anyi