Data Science

Data science studies the extraction of patterns from large, often unstructured datasets: statistical models, machine learning and the methodological question of whether a found pattern is genuinely causal or merely coincidentally correlated. The field supplies the analytical method; the application domain — energy, maintenance, cost — determines what the data means.

FM by now collects large volumes of use, condition and cost data, but rarely has the statistical literacy to interpret it responsibly. Without data science knowledge, correlation is too quickly taken for causation, and a model built on too little data is treated as a reliable prediction.

• Is the dataset large and representative enough for the model built on it?
• Does this model explain something, or does it only predict — and does the user know the difference?

• Statistical and machine-learning literature; academic research applying data science to building data (e.g. energy forecasting, failure prediction).

• Every predictive model in FM should come with a confidence margin, not a single figure presented as certainty.

• CAFM & Information Management (data source)

• Data analysis and model interpretation

• Demand organisation

  Data science capacity is rarely built within FM itself; the client role must decide whether this is bought in or shared with a central data function.

• No ISO standard specific to data science application in FM; methodological quality rests on general statistical practice.

The standards section arrives in Part 7.

• A model trained on one building presented as generically applicable across the whole portfolio.

Explainable AI for building applications is still a young research field; most usable predictive models in FM remain black boxes whose output users must trust without being able to follow the reasoning.

• Academic journals on applied data science in the built environment, e.g. Building and Environment.