There’s buzz everywhere these days about exciting applications of data science, especially related to machine learning. While the majority of the media attention focuses on big names and sexy topics — e.g., the Google Brain, image classification, speech recognition — we are fortunate to work with companies every day that are doing fascinating work with data science tools. Although sometimes more mundane, many of their use cases have more real-world impact.
To illuminate the diversity of domains where data science can be applied, I want to share some of the more interesting applications we’ve run across. Unfortunately, I can’t elaborate on some of these for confidentiality reasons — but I hope they spark your imagination and creativity. What new idea will they inspire for you?
Kat Risk developers high-precision models of risk from weather-related catastrophes (e.g., flooding, wind damage). They have some free demos you can [check out on their website].
Predictive car maintenance
A major car company exploring how deep learning can react to audio recordings from the engine to determine if maintenance is necessary, or if parts are nearing the need for replacement.
Mental health care
Ginger.io uses data from users’ mobile devices to form a view on how users are feeling. This is useful for the end users, as well as their doctors. According to Ginger’s website, their “behavioral
analytics engine, built from years of research at the MIT Media lab, aggregates, encrypts, and anonymizes patient data before running it through statistical analysis to create meaningful insights.”
Numenta has released Grok, an application that monitors servers and alerts you about anomalies. “Through complex pattern analysis, Grok identifies abnormal conditions or gradual trends – situations that tools based on thresholds or simple statistics can easily miss.”
Air travel claims
European regulations require airlines to compensate passengers for delayed or canceled flight, and Flightright helps customers recover money by handling the bureaucracy of the process for you. They use data science to predict the likelihood of your claim being successful, based on the information you provide
A company is using data science to make assessments about the risk associated with lending money for bail, including the risk that a defendant will violate his or terms of release.
Case Commons is a non-profit that makes software to help social workers and administrators of social welfare programs. According to their website, their software uses “predictive algorithms that can, for example, show a given child’s projected path. This means that we can enrich views of information for managing current caseloads with what agency history tells us about longer-term outcomes in areas such as permanency, education and health.”
Claims adjusters for car accidents
An automotive insurance company is using data science to automate some of the work of their claims adjusters. Instead of spending lots of time sending adjusters out to put together an assessment of an accident, they are using data about the accident to predict the necessary repair costs. This won’t completely replace human adjusters, but it will make them much more efficient.
[Quantitative Analytics] works in conjunction with a number of agencies, including NOAA, to improve wildlife management in the Pacific Northwest. For example, one of their projects involves modeling the effects of habitat restoration on salmon population, so they can optimally direct habitat restoration efforts.
Donor Bureau uses data science to improve targeting for direct mail campaigns. They work with non-profits and political groups to maximizing return for each letter sent in direct mail campaigns. Their data scientists train targeting models against a data warehouse of over a billion transactions and tens of millions of donors.