AI case studies from the real-world

Neopost identifies customers at risk of churn with machine learning using PredicSis

CargoMetrics analyses satellite shipping imagery and plans to to identify investment opportunities based on the data with machine learning

Portcast is a data analytics vendor promising to optimise logistics for the maritime industry using machine learning

Rolls Royce plans to predict maintenance requirements for jet engines to improve aircraft efficiency using Microsoft Azure's machine learning

Transavia aims at reducing its 2018 carbon footprint by 6500 tonnes using machine learning

France Telecom's Telekomunikacja Polksa realised that certain customers have a greater or lesser influences on networks of mobile phones users. If highly connected networkers churn then this is likely to cause a large ripple effect. To improve customer churn prediction and identification of who to retain they developed social graphs and analysis based on the transaction history and network connections of customers. This allowed them to improve prediction by 47%.

US cellular increases customer lifetime value by 61% by analysing website visitor behavior against online and offline sales, subscriptions and renewals using machine learning

H2O.ai to offer interpretable ready-to-deploy deep learning and machine learning algorithms for enterprises to enable a faster transition to AI

IBM launches Equipment Maintenance Assistant to predict equipment maintenance requirements and aid technicians in diagnosis and repair using machine learning

Lobe is a visual tool that can be used to build custom deep learning models without writing any code

Microsoft trials AI to analyse strategic partners contracts

US Army test computer vision for an enemy infrastructure recognition system in real-time

The London Fire Brigade identifies the areas prone to fires in ducting systems to provide targeted maintenance using natural language processing

US FEMA estimates building damage caused by volcano eruptions using satellite imagery and convolutional neural networks

ICRISAT and Microsoft enable Indian farmers to achieve up to 30% higher yields at trial through machine learning advice on when to sow crops

Chinese Government hopes to improve decision support to diplomats working on Bridge and Belt initiative through investment in AI tools

City of Redwood City improves parking occupancy utilisation by 250% and optimises revenue using machine vision

New York is estimating street volumes and understanding how street design can reduce injuries using predictive algorithms

Seattle is identifying collision patterns and factors that contribute to higher levels of injury severity

Swiss Government tests allocating immigrants to appropriate geographic areas in Switzerland

The Government of Canada plans to algorithmically scan the internet for data on public perceptions of cannabis use and related behaviours

The Northern Territory Department of Primary Industry and Resources achieves 95% accuracy identifying fish with machine learning

The US National Oceanic and Atmospheric Administration is using AI to improve real-time decision-making for high-impact weather

The US Veterans Affairs was able to increase therapy completion rate among war veterans with PTSD by more than six times using machine learning to personalise the program

The World Bank is exploring using machine learning to analyse and improve public spending processes and find signals indicating potential corruption with Microsoft Research

Boston Public Schools's plan to reconfigure start times for high school students using an optimisation algorithm backfires

Harvard university seismologists developed a deep learning algorithm to predict the location of aftershocks with an AUC of .849

Monta Vista High School students build a model to predict wildfires using Google Tensorflow

Peptone accelerates protein research for drug discovery with a machine learning derived database

Researchers at Macau University of Science and Technology develop a new model for disease classification in cases of limited labelled data with ~90% accuracy by combining logistic regression and semi-supervised learning