CTO and Founder, SpaceCurve
Prediction is the process of taking a measurement of reality and contextualizing it around a model of how the world works. Building models that attempt to approach the richness of the real world not only requires big data, it also requires working with data from diverse sources that capture the many aspects of reality.
The challenge, especially for next generation big data, is when we start talking about ‘data fusion,’ or the joining of many diverse data sources such as social media, satellite imaging, mobile phone telemetry, and weather sensors. These sources were never designed to work together. We are tasked with finding the primary keys to join all of these data models together into a single, unified model that exposes the relationships between unrelated data elements.
Space and time are interesting attributes of data and events because they represent a fundamental organizing principle of reality that constrains causality. In essence, the combination of space and time is the primary key of reality. Additionally, most data sources have space and time as a meaningful attribute, thereby introducing a commonality across previously disparate data sets.
This is an incredibly powerful mechanism by which we can build unified views of reality that we weren’t able to build before.
Traditionally, we would take a more siloed approach, where the data focus is on individual data source such as social media, geospatial data, video, etc. The problem with silos for predictive analytics is that they only capture a narrow sliver of the available context. When you start taking all of the potential sources and context around what is happening at any point in time to build models of behavior, then leveraging space and time relationships becomes one of the most fundamental organizing principles for predictive analytics.
If you’d like to hear more from Andrew on the topics of big data, predictive analytics and SpaceCurve, join us Wednesday evening, May 14 for our Evening Forum program Big Data, Small Things, Predictive Analytics.