Smart applications we increasingly see in Cities, factories and homes are largely driven by data, thus using some intelligence to analyze and correlate those pieces of information. Data analytics methods have advanced over time and predictive capabilities have become more elaborate, with Artificial Intelligence (AI) now playing a major role in supporting data comprehension and related decision making.
AI allows data analysis to evolve from forecasting to truly predictive models. While forecasting is the process of calculating future events or trends by leveraging past and present data, AI predictive capabilities use data mining and enhanced algorithms to estimate more granular, specific outcomes, making assumptions and continuously learning at a large scale and incredible depth of detail.
For example, forecasting techniques are typically applied to business processes to assess future sales upon a mix of variables such as past results, upcoming market opportunities, customer insights, marketing activities, etc. AI predictive capabilities are nowadays used in a number of domains, from weather forecast to environmental protection, up to public health and epidemic studies.
How much accurate are AI-generated predictions? Some recent pilot cases make AI seem quite mature and trustworthy.
The City of London in Ontario, Canada is implementing the Chronic Homelessness Artificial Intelligence model to predict and prevent homelessness, using machine learning to calculate the probability of an individual in the city’s shelter system becoming chronically homeless within the next six months. During the testing phase, the model proved a 93% accuracy.
In Australia AI is being used to better understand and predict bushfires, while an American high school student developed an AI framework to anticipate future air pollution levels with 92% accuracy. He said he wanted to give his grandparents the opportunity to spend some time outside in safety by seeing air quality a few days in advance, so he combined multiple machine learning techniques to process publicly available weather and air pollution data to get reliable information.
AI predictive capabilities could also improve the way physicians, healthcare providers and patients interact with health data. Promising models are being tested to predict diabetics with highest risk of avoidable complications, hopefully reducing hospitalizations and readmissions, or to support cardiology monitoring by identifying problems such as sudden cardiac death and MI before they occur.
Since AI and its predictive capability are making great strides, we shouldn’t forget that predictions can only be as good as the input data – so the way we collect, transport, store, share, and protect the ever-increasing volume of private and public information could really impact the quality of present and future data-driven decision.